Initialize portable humanoid RL framework with licenses, locks and CI
Core and layout checks / core (push) Has been cancelled

This commit is contained in:
2026-11-03 19:46:08 +08:00
commit 7e1ff58b46
436 changed files with 64144 additions and 0 deletions
+32
View File
@@ -0,0 +1,32 @@
---
name: controls
description: Inspect simulation observation/action contracts, joint ordering, PD/effort settings and measured joint replay without enabling physical robot control.
---
# Inspect controls and measured replay
This manual concerns **simulation/inference controls**, not a ROS/DDS or motor-bus implementation.
Read [Robot checks](../../../docs/NEW_ROBOT.md), [Training operations](../../../docs/TRAINING_OPERATIONS.md)
and [Deployment contract](../../../docs/DEPLOYMENT.md).
## Procedure
1. Re-run `./scripts/uh check-robot --config FILE --output NEW_REPORT` after changes to joints,
gains, torque limits, sensors, reset pose or model assets. Inspect resolved order, not regex order.
2. Distinguish observation joint order from action order; check frames, radians/rad/s, action
scaling/offset/clipping, physics dt, policy dt and normalizer ownership. Never normalize twice.
3. Record `./scripts/uh play --checkpoint CHECKPOINT --measure FILE.csv --steps 1000 --profile stress`.
Use `--video FILE.mp4` for visual context. Repeat randomized clean-observation evaluation to
separate observation-noise effects. Interactive `play --viewer native|viser` is a different mode.
4. Inspect joint positions, velocities, projected actuator torque, raw/applied actions, targets and
terminal flags. Measurements occur after each control step BEFORE reset; torque is an endpoint
actuator contribution, not a desired torque, substep peak or physical sensor measurement.
5. Look for saturation, large action oscillations, limit proximity, sliding and collapsed posture.
State the evidence; do not silently change gain/clip settings to make the replay appear better.
## Boundaries and handoff
A sensor named IMU does not validate a real estimator. Simulated PD gains are not a hardware tuning
recipe. There is no joystick-to-motor path, bus mapping, E-stop or autonomous robot operation in this
repository. Provide contract/check/report/CSV/video paths and delegate actual hardware integration
only after explicit operator authorization and a separately designed safety system.
+32
View File
@@ -0,0 +1,32 @@
---
name: deploy
description: Export and validate research-only ONNX inference bundles with matching evaluation provenance, numerical parity and explicit hardware exclusions.
---
# Package a policy for research inference
Read [Deployment contract](../../../docs/DEPLOYMENT.md). Complete train/controls evidence first.
This is not a hardware deployment skill: no motor enable, robot transport or firmware operation.
## Procedure
1. Select a committed checkpoint from a completed run. Inspect its training curves, conformance
report, fixed-protocol evaluation, censoring, CSV and video. Export can package a poor policy;
it does not certify quality.
2. Finish baseline evaluation/export before editing source. Restore historical sources/dependencies
in a separate trusted environment when needed; do not bypass provenance checks.
3. Run `./scripts/uh export --checkpoint CHECKPOINT --evaluation REPORT --output NEW_BUNDLE`.
The evaluator must match checkpoint, experiment and source identity. ONNX parity uses actual
rollout observations, not just random input vectors.
4. Run `./scripts/uh verify-bundle BUNDLE --inference`. Inspect observation fields, normalizer,
joint order, action scale/offset/clip, PD metadata, units and control timing.
5. Review asset/code licenses and manifests for private paths before publishing artifacts.
Source archives are trusted restoration material, not automatically executable downloads.
## Failure handling and handoff
Hash mismatch, source drift, unsupported action schema or failed parity is a blocker. Do not
edit hashes or tolerance to turn an unknown failure into success. Hand over bundle/report paths,
measured parity, training/evaluation limitations and the fact that `hardware_approved` remains
false. Physical use additionally needs validated estimation, transport, joint mapping, watchdog,
limits, emergency stop and operator-approved testing; those are absent here.
+39
View File
@@ -0,0 +1,39 @@
---
name: new-task
description: Add a robot or task with explicit mechanics, independent conformance declarations, negative tests and bounded simulation acceptance.
---
# Add a robot or task
Read [Robot checks](../../../docs/NEW_ROBOT.md) and [Architecture](../../../docs/ARCHITECTURE.md).
Use setup-env first if runtime origins or CUDA are unverified.
## Procedure
1. Identify the exact model variant and asset license. Run `./scripts/uh inspect-asset PATH`.
Compilation does not validate sensor frames, contact semantics or control assumptions.
2. Implement independent robot/environment/PPO factories. Establish foot sites, IMU placement,
collision filters, gains/effort limits, joint order and nominal posture from reviewed sources.
Do not derive motor ratings from names or copy another robot's gains without justification.
3. Register the task in `src/universal_humanoid/registry.py`; create a bounded experiment JSON.
Keep the CLI robot-independent and do not change an existing task ID's mechanics silently.
4. Add `src/universal_humanoid/robot_specs/<robot>.json` with independently reviewed expected values.
It is a contract, not an auto-blessed dump of the current model.
5. Run `./scripts/uh check-robot --config FILE --output NEW_REPORT` and read every failed check ID.
Add negative tests for a missing site, wrong IMU frame, bad PD gain, swapped order and lost contact.
6. Run `./scripts/check`, zero-action smoke, short PPO, randomized/stress evaluation, measured
replay and export parity. Record all output paths and failures.
## Boundaries
Current conformance/export assumes flat-ground velocity tracking and one position actuator per
controlled hinge. Extend and version the contract for coupled drives or new task semantics;
do not disable checks to fit a different robot. A passing check means adapter conformance, not
walking ability, mechanical accuracy or hardware approval.
## Handoff
Supply factories, registry/spec/experiment paths, asset attribution, conformance report, negative
tests and measured pipeline evidence. State which properties remain unvalidated. Changes to
vendored code additionally require the vendor-review procedure; routine new adapters should live
in the first-party package.
+105
View File
@@ -0,0 +1,105 @@
---
name: rl-engineer
description: Engineer humanoid reward revisions, bounded mjlab training, measured evaluation, replay and research-only ONNX packaging in Universal Humanoid.
---
# RL engineer
## 1. Establish a contract
Use this as the cross-stage coordinator; consult setup-env, new-task, train, controls and deploy
through the manual index in `AGENTS.md` for stage-specific work.
Read `AGENTS.md`, `README.md`, and the target experiment. Confirm the robot/task, desired
behavior, resource budget and acceptance metrics. A request to build architecture authorizes
short smoke validation, not days of training or physical robot operation. Use the current coding
agent directly; do not launch subagents without operator authorization.
Run `./scripts/uh doctor` and `list`. Inspect the factory's source and relevant mjlab reward
functions before editing. Follow `docs/SETUP.md` to verify the installed vendored backends,
locked dependencies and bundled assets. Use `verify-vendor` to check the source inventory.
Do not install a second mjlab/rsl-rl-lib distribution. Record actual dependency provenance.
## 2. Baseline and reward hypothesis
Copy a baseline experiment into a NEW JSON file with an explicit task and seed. Write the
hypothesis next to the experiment or in a run note. Inspect `uh rewards --config FILE`.
- Weights/parameters: edit JSON; misspelled term names must fail.
- New mechanisms: add vectorized functions to `src/universal_humanoid/rewards.py` and the
`FUNCTIONS` allowlist; add tests for shape `[num_envs]`, finiteness, sign and limiting cases.
- Do not change observations/control/mechanics while claiming a reward-only comparison.
- Never `eval` code supplied in experiment JSON.
Run baseline evaluation/export before editing Python. Source hashes are strict; archived sources
may be restored only into a separate trusted workspace, never over a shared reference checkout.
## 3. Cheap checks before expensive runs
Run `./scripts/check`, `uh plan`, `uh rewards`, and `uh check-robot --config FILE --output NEW_REPORT`,
then `uh smoke` with 4 environments and 16 steps.
For each new robot, verify free root, DoF, mass/inertia, mesh resolution, contacts, foot sites,
IMU frame, torque limits, gains, action ordering, reset posture and observation terms. Model
compilation alone cannot validate these semantics. Zero-action success is not balance training.
## 4. Bounded training
Use `uh train --config FILE --output NEW_DIRECTORY`, or `pipeline` for a smoke-sized end-to-end
run. Read `docs/TRAINING_OPERATIONS.md` for lifecycle and statistics semantics. Use
`train --background` for a requested long-running Linux job, `status RUN` to query progress,
`checkpoints RUN` to inspect committed checkpoints, and `stop RUN` for a guarded graceful stop.
Resume with `train --resume CHECKPOINT --iterations ADDITIONAL --output NEW_DIRECTORY`;
optimizer continuation is not bitwise simulator/RNG replay. Never guess checkpoints by mtime.
Keep TensorBoard local; no W&B authentication or automatic upload. Use `curves RUN --output NEW_DIR`
to inspect tracking, action distribution std and episode-length trends DURING training. Report
missing tags, non-finite samples and short curves. Rising total reward alone is not learning.
Inspect `run.json`, `progress.json`, `resolved_env.yaml`, `resolved_agent.json` and `sources.zip`.
Long runs require an explicit operator-approved budget. Do not automatically increase parallelism
past measured VRAM capacity. This backend requires CUDA; do not pretend Warp CPU is fast CPU RL.
## 5. Evaluate and replay
Use `benchmark --checkpoints ... --seeds ... --profiles randomized stress --output NEW_DIR`
for stand `(0,0,0)`, forward `(0.3,0,0)`, lateral `(0,0.2,0)`, turn `(0,0,0.3)`. Keep identical
seeds, profiles, world counts and horizons for candidates. These are evaluation seeds, NOT
independent training seeds; separately train multiple seeds when investigating training variance.
Default evaluation randomizes each world's reset state. `stress` also enables actor observation
noise and per-world interval velocity kicks; inspect actual initial states and push counts.
`nominal` intentionally repeats fixed initial states. Startup mass/friction randomization is not
part of these profiles. None constitutes full sim-to-real stress coverage.
Fall/time-limit terminations remain active; curriculum is disabled. Report RMSE, completed-only
failure fraction, failure events per simulated world-second, first-episode failures AND censoring.
No completed episodes means completed-only means/fractions are null. Variance is across seed-level
reports, not repeated worlds, and is null for fewer than two values. Prefer horizons longer than
the episode duration. `compare` supports paired matrices and refuses protocol mismatches.
Use `select-checkpoint` with an explicit metric, inspect trade-offs, then evaluate held-out seeds;
never equate a selection-set winner with generalization or automatically promote it to hardware.
Record MP4 with `evaluate --video FILE`. Use `play --measure FILE.csv --steps N` for bounded
headless joint/velocity/actuator-torque measurement before terminal-world reset, or
`play --viewer viser` for interactive inspection. CSV samples are policy-rate endpoints, not
substep torque peaks. Refer to the CSV sidecar for units and completion status.
Look for reward hacking: standing instead of tracking, sliding, knee collapse, asymmetric contacts,
large actions, unrealistic high-frequency gait. Videos are complementary to metrics, not proof.
## 6. Package, never auto-enable hardware
`uh export --checkpoint ... --evaluation ... --output NEW_DIRECTORY` checks report identity and
ONNX parity on real observations. `uh verify-bundle DIRECTORY --inference` rechecks checksums
and the saved reference. Output is a research bundle even when rollout succeeds. Inspect the
resolved action joint order, offsets/scales, observation ordering and simulation motor gains.
Physical deployment requires a separate validated adapter for transport, estimator, bus order,
calibration, bounds, timeout watchdog, torque ramp, E-stop and operator authorization. There is
no motor-enable command in this repository. Never change `hardware_approved` to bypass this.
## 7. Handoff
Report checkpoint, backend, env/iteration counts, tracking/std/episode curves, evaluation matrix,
CSV/video paths and measured evidence—not only wall time or command exit codes. State whether learning was
actually demonstrated. List limitations, failed attempts and next validation requirements. Avoid
claiming general morphology support from the fact that two task adapters share a CLI.
+37
View File
@@ -0,0 +1,37 @@
---
name: setup-env
description: Detect the host, install the locked vendored runtime in a fresh environment, and verify dependencies, backend origins, CUDA and robot gates.
---
# Set up and verify an environment
Use for first installation, another machine, missing imports, wrong backend, or CUDA failures.
Read [Setup](../../../docs/SETUP.md) and [Vendor provenance](../../../docs/VENDOR.md).
## Procedure
1. Run `python3 .agents/skills/setup-env/scripts/detect.py`. Record Python/platform, GPU/driver,
installer availability and existing import origins. Detection is not a simulation pass.
2. Confirm Linux x86-64 and compatible NVIDIA CUDA support before choosing the reference lock.
Do not substitute CPU training or silently change Torch/Warp versions.
3. Run `uv sync --locked --extra dev --python 3.13` in a fresh project environment. The ordinary
`pip install -e .` path and hashed pip alternative are documented in Setup. Never install
separate `mjlab`/`rsl-rl-lib` copies into this environment or repair another project's venv.
4. Run `./scripts/uh doctor`, `./scripts/uh verify-vendor`, and `./scripts/check`.
5. Run `.venv/bin/python .agents/skills/setup-env/scripts/check.py --config experiments/g1_smoke.json --output runs/setup-g1`.
Use the T1 smoke config for T1. Read every gate log and `summary.json`, not just a GPU name.
## Failure handling
- A source mismatch means wrong installation or an unreviewed vendor change. Inspect it; do not
refresh hashes or unset errors to manufacture success.
- A locked wheel download can be slow. Report installation as pending until the installer exits
successfully and dependency checks pass.
- A robot check failure belongs to the new-task procedure; changing rewards cannot repair an IMU frame.
- An unavailable GPU/driver is an external prerequisite, not permission to claim CPU equivalence.
## Handoff
Report interpreter, installed backend origins, lock/inventory status, GPU/driver, gate output
paths and remaining failures. Only then authorize a bounded training smoke within the operator's
budget. No credential collection, remote installation or hardware operations are part of setup.
+56
View File
@@ -0,0 +1,56 @@
#!/usr/bin/env python3
"""Installed-environment gates with durable command logs; never installs or repairs packages."""
import argparse
import json
import shutil
import subprocess
import sys
from pathlib import Path
def main():
parser = argparse.ArgumentParser(description=__doc__)
parser.add_argument("--output", type=Path, required=True)
parser.add_argument(
"--config", type=Path, help="Also check this robot adapter and run a bounded GPU smoke"
)
args = parser.parse_args()
args.output.mkdir(parents=True, exist_ok=False)
base = [sys.executable, "-m", "universal_humanoid"]
commands = [("vendor", [*base, "verify-vendor"]), ("runtime", [*base, "doctor"])]
if shutil.which("uv"):
commands.append(("dependencies", ["uv", "pip", "check", "--python", sys.executable]))
else:
commands.append(("dependencies", [sys.executable, "-m", "pip", "check"]))
if args.config:
commands += [
(
"robot",
[
*base,
"check-robot",
"--config",
str(args.config),
"--output",
str(args.output / "robot.json"),
],
),
(
"smoke",
[*base, "smoke", "--config", str(args.config), "--steps", "8", "--num-envs", "4"],
),
]
results = []
for name, command in commands:
with (args.output / f"{name}.log").open("x") as stream:
result = subprocess.run(command, stdout=stream, stderr=subprocess.STDOUT, check=False)
results.append({"gate": name, "command": command, "returncode": result.returncode})
report = {"ready": all(r["returncode"] == 0 for r in results), "gates": results}
(args.output / "summary.json").write_text(json.dumps(report, indent=2) + "\n")
print(json.dumps(report, indent=2))
raise SystemExit(0 if report["ready"] else 1)
if __name__ == "__main__":
main()
@@ -0,0 +1,56 @@
#!/usr/bin/env python3
"""Standard-library-only host report; safe to run before installing the project."""
import argparse
import importlib.util
import json
import platform
import shutil
import subprocess
import sys
from pathlib import Path
def main():
parser = argparse.ArgumentParser(description=__doc__)
parser.add_argument("--output", type=Path)
args = parser.parse_args()
nvidia = shutil.which("nvidia-smi")
gpu = (
subprocess.run(
[nvidia, "--query-gpu=name,driver_version,memory.total", "--format=csv,noheader"],
capture_output=True,
text=True,
timeout=15,
)
if nvidia
else None
)
origins = {}
for name in ("universal_humanoid", "mjlab", "rsl_rl", "torch"):
spec = importlib.util.find_spec(name)
origins[name] = spec.origin if spec else None
report = {
"python": sys.executable,
"version": platform.python_version(),
"platform": platform.platform(),
"machine": platform.machine(),
"uv": shutil.which("uv"),
"nvidia_smi": nvidia,
"gpu_query": gpu.stdout.strip() if gpu else None,
"gpu_query_error": gpu.stderr.strip() if gpu and gpu.returncode else None,
"import_origins": origins,
"note": (
"Discovery only, not a dependency/CUDA/simulation pass. "
"Run setup-env check.py after installation."
),
}
if args.output:
args.output.parent.mkdir(parents=True, exist_ok=True)
with args.output.open("x") as stream:
json.dump(report, stream, indent=2)
print(json.dumps(report, indent=2))
if __name__ == "__main__":
main()
+39
View File
@@ -0,0 +1,39 @@
---
name: train
description: Operate bounded PPO experiments, inspect learning curves, resume committed checkpoints and compare measured multi-seed evaluations rather than elapsed time.
---
# Train and measure a policy
A finished run is not a learned policy. Read [Training operations](../../../docs/TRAINING_OPERATIONS.md)
and the reward-experiment sections of [rl-engineer](../rl-engineer/SKILL.md).
## Procedure
1. Confirm setup-env and the target robot gate passed. Agree on task, resource budget and acceptance
metrics. Measure GPU capacity before increasing worlds; do not launch the long default automatically.
2. Inspect `./scripts/uh plan --config FILE` and `./scripts/uh rewards --config FILE`; make a named
reward hypothesis without silently changing observations, controls or robot mechanics.
3. Smoke first, then `./scripts/uh train --config FILE --output NEW_RUN --background` when requested.
Query `./scripts/uh status RUN` and read `./scripts/uh curves RUN --output NEW_ANALYSIS` during training.
Inspect tracking, action std, episode length and losses together. Report missing or short curves.
4. Use `./scripts/uh checkpoints RUN`; only committed manifest entries are eligible. Stop with
`./scripts/uh stop RUN`. Resume explicitly with `train --resume CHECKPOINT --iterations ADDITIONAL`
and a new output directory. Never rewrite a failed run to completed or promise bitwise replay.
5. Run `./scripts/uh benchmark --checkpoints ... --seeds ... --profiles randomized stress --output NEW_DIR`.
Evaluate standing/forward/lateral/turning, keep protocols identical, and inspect censoring and actual pushes.
6. Use explicit-metric `select-checkpoint` and held-out evaluation seeds, then the controls manual
for measured joint replay and deploy for research packaging.
## Interpret evidence
Rising total reward with flat tracking can indicate reward exploitation. More time or updates do
not guarantee improvement. Evaluation seeds are not independent training replicas; a training-seed
study needs separately trained policies. Sample variance with one seed is undefined, not zero.
Compare task metrics, not raw totals from different reward functions.
## Handoff
Report checkpoint/backend/env/update counts, tracking/std/episode curves, evaluation table,
failure-rate definitions, CSV/video and remaining limitations. Do not declare convergence from a
smoke run or approve physical motion from an ONNX file.
+9
View File
@@ -0,0 +1,9 @@
* text=auto eol=lf
# Vendor inventories fingerprint exact upstream bytes, including original line endings.
rl/** -text
*.stl binary
*.STL binary
# A blank context line in an archived unified diff is a literal space; preserve it.
rl/upstream/*.patch -whitespace
+46
View File
@@ -0,0 +1,46 @@
name: Core and layout checks
on:
push:
pull_request:
workflow_dispatch:
permissions:
contents: read
jobs:
core:
runs-on: ubuntu-latest
timeout-minutes: 15
env:
PYTEST_DISABLE_PLUGIN_AUTOLOAD: '1'
steps:
# Immutable revisions also used by the Jumper reference workflow.
- uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
- uses: actions/setup-python@a26af69be951a213d495a4c3e4e4022e16d87065
with:
python-version: '3.13'
- name: Install pinned resolver
run: python -m pip install uv==0.13.1
- name: Verify dependency lock and exports
run: |
uv lock --check
uv export --locked --extra dev --no-emit-project --format requirements-txt --output-file requirements-linux-cu130.lock
uv export --locked --only-group ci --no-emit-project --format requirements-txt --output-file requirements-ci.lock
git diff --exit-code -- uv.lock requirements-linux-cu130.lock requirements-ci.lock
# Core/static tests only: this group installs no project, Torch or CUDA stack.
- name: Install locked check and build tools
run: |
uv sync --locked --only-group ci --python 3.13
uv pip install --python .venv/bin/python --no-deps --no-build-isolation -e .
- name: Check vendor inventory and wheel resources
run: |
.venv/bin/python tools/vendor_manifest.py
.venv/bin/python -m build --wheel --no-isolation
.venv/bin/python tools/check_wheel.py dist/universal_humanoid-*.whl
- name: Lint and format
run: |
.venv/bin/python -m ruff check src tests tools .agents/skills/*/scripts
.venv/bin/python -m ruff format --check src tests tools .agents/skills/*/scripts
- name: Core and layout tests
run: .venv/bin/python -m pytest -q
+27
View File
@@ -0,0 +1,27 @@
.venv/
.venv-*/
build/
dist/
__pycache__/
*.py[cod]
*.egg-info/
.pytest_cache/
.ruff_cache/
runs/
bundles/
analysis/
*.job/
*.mp4
*.pt
*.onnx
.env
.env.*
!.env.example
*.pem
*.key
logs/
out/
.coverage
htmlcov/
.DS_Store
MUJOCO_LOG.TXT
+59
View File
@@ -0,0 +1,59 @@
# Agent engineering contract
Keep project changes within this repository. Treat external dependencies and robot asset
checkouts as read-only unless the operator explicitly authorizes changes to them.
## Task manuals
Read the matching manual before acting; this index serves agents that do not auto-discover skills.
| Manual | Use when |
|---|---|
| [setup-env](.agents/skills/setup-env/SKILL.md) | Detect/install a new environment or diagnose backend/CUDA dependencies |
| [new-task](.agents/skills/new-task/SKILL.md) | Add a robot/task and run mechanical/interface conformance checks |
| [train](.agents/skills/train/SKILL.md) | Train, inspect learning curves, manage checkpoints/jobs and compare experiments |
| [controls](.agents/skills/controls/SKILL.md) | Inspect control contracts, joint/torque CSV and simulation replay |
| [deploy](.agents/skills/deploy/SKILL.md) | Export/verify research bundles and state hardware exclusions |
| [rl-engineer](.agents/skills/rl-engineer/SKILL.md) | Coordinate a reward-to-evaluation experiment across the stages |
Detailed references: [Setup](docs/SETUP.md), [Robot checks](docs/NEW_ROBOT.md),
[Training operations](docs/TRAINING_OPERATIONS.md), [Vendor sources](docs/VENDOR.md),
[CLI reference](docs/CLI.md). `tests/test_layout.py` enforces index/link/CLI synchronization.
## Boundaries
- The coding agent is the orchestrator; use `scripts/uh`. No separate LLM service is required.
- Do not delegate work or start multi-agent workflows unless the operator asks for them.
- Use installed vendored backends in `rl/`; do not install separate mjlab/rsl-rl-lib distributions.
- Verify `doctor`, `verify-vendor` and the lock before changing dependencies; no personal path fallbacks.
- Vendor changes require explicit review, modification notices and inventory updates; never refresh hashes merely to silence failures.
- Public documentation must distinguish reusable features from reference-environment evidence and release blockers.
- Never change the task's robot, joint order, actuator model or observation dimensions silently.
- Keep hardware facts in robot adapters, reward mechanisms in `rewards.py`, and tuning in experiments.
- Never disable falls to manufacture a successful evaluation. Keep finite episodes and censoring counts.
- Never call a two-iteration run a converged policy, nor call an ONNX bundle hardware-ready.
- No physical robot bus, motor enable, firmware flash or autonomous real-world deployment.
- Do not run untrusted checkpoints: the upstream loader uses pickle.
- Do not delete failed runs. New experiment, new directory. Read failure status and logs.
- Evaluate and export a baseline before editing source. Provenance intentionally rejects source drift.
- Read training curves (tracking, action std, episode length); elapsed time is not evidence of learning.
- Use explicit checkpoint manifests, status/stop and new-directory resumes; never guess the newest file.
- Distinguish evaluation seeds from training replicas and completed-only failure fractions from incidence.
- Check actual randomized states, push counts and censoring before claiming stress-test coverage.
## Checks
`./scripts/check` runs lint and tests without inherited ROS pytest plugins.
`./scripts/uh check-robot --config experiments/<robot>_smoke.json --output NEW_REPORT` checks
sites, IMU, contacts, PD/limits, joint order and initial posture against declared expectations.
`./scripts/uh smoke --config experiments/<robot>_smoke.json` checks actual GPU dynamics.
A full acceptance check needs bounded PPO training, a fixed-command evaluation and ONNX parity.
A task is supported only after its sensors/collisions, resolved action order, reward finiteness,
training, evaluation and export have been checked. Asset discovery alone is not support.
## Reporting
State exact config/run/checkpoint/report paths, GPU, iteration/env/step counts and measured metrics.
Separate plumbing validation, learning quality, export parity and physical safety. Never select a
reward revision by raw reward total alone; compare fixed-protocol tracking and failures across seeds.
+9
View File
@@ -0,0 +1,9 @@
# Repository instructions
Read [AGENTS.md](AGENTS.md) first. Its task-manual index, safety boundaries and reporting rules
apply to this repository. Use the matching manual rather than assuming a task can be solved
by editing rewards alone.
The simulator and PPO sources are installed from this repository's `rl/`; do not select an
external backend through PYTHONPATH or modify another project's environment. Setup and robot
conformance gates precede training. No command in this project authorizes physical robot motion.
+36
View File
@@ -0,0 +1,36 @@
# Contributing
Read [AGENTS.md](AGENTS.md) for the task-manual index and safety boundaries. Install the locked
runtime using [Setup](docs/SETUP.md), then run `./scripts/check`. The GitHub core/layout job does
not provision CUDA and is not a substitute for the GPU setup/robot/pipeline gates.
## Where changes belong
- `src/universal_humanoid/`: first-party CLI, task adaptation, evaluation and operations.
- `src/universal_humanoid/robot_specs/`: independently reviewed robot conformance expectations.
- `experiments/`: bounded, explicit experiment recipes; no automatically launched long training.
- `rl/mjlab/`, `rl/rsl_rl/`: reviewed vendored dependencies, not a place to scatter routine task changes.
- `.agents/skills/`: stage manuals; index every new manual in `AGENTS.md`.
- `docs/`: public setup, architecture, interface contracts and validation records.
- `tests/`: core, adapter, negative conformance, provenance, telemetry and layout regression checks.
## Required checks
`tests/test_layout.py` requires the CLI table to match parser commands/options, robot check IDs
to match their documentation, every skill to be indexed, links to resolve, backend packages and
licenses to be present, and vendor inventory/portable path rules to hold. Update code and its
manual together; do not weaken a test just to avoid maintaining documentation.
For a robot change, follow [NEW_ROBOT.md](docs/NEW_ROBOT.md), including negative tests and actual
contact/control probes. For a backend change, follow [VENDOR.md](docs/VENDOR.md), retain modification
notices and licenses, review inherited patches, and regenerate the vendor inventory deliberately.
Dependency changes require a reviewed `uv.lock` and refreshed hashed pip export.
Report the exact configuration, source/asset identity, runtime, commands, outputs and observed
failures. Distinguish static/unit success, physics/interface conformance, policy quality and ONNX
parity. Never claim hardware safety from simulation or submit credentials, environments, logs,
private asset files or unreviewed source archives.
Original project code and documentation use [Apache-2.0](LICENSE). Intentionally submitted
contributions follow its contribution terms unless explicitly stated otherwise. Retain third-party
copyright/license notices and do not relicense bundled dependencies or assets.
+202
View File
@@ -0,0 +1,202 @@
Apache License
Version 2.0, January 2004
http://www.apache.org/licenses/
TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION
1. Definitions.
"License" shall mean the terms and conditions for use, reproduction,
and distribution as defined by Sections 1 through 9 of this document.
"Licensor" shall mean the copyright owner or entity authorized by
the copyright owner that is granting the License.
"Legal Entity" shall mean the union of the acting entity and all
other entities that control, are controlled by, or are under common
control with that entity. For the purposes of this definition,
"control" means (i) the power, direct or indirect, to cause the
direction or management of such entity, whether by contract or
otherwise, or (ii) ownership of fifty percent (50%) or more of the
outstanding shares, or (iii) beneficial ownership of such entity.
"You" (or "Your") shall mean an individual or Legal Entity
exercising permissions granted by this License.
"Source" form shall mean the preferred form for making modifications,
including but not limited to software source code, documentation
source, and configuration files.
"Object" form shall mean any form resulting from mechanical
transformation or translation of a Source form, including but
not limited to compiled object code, generated documentation,
and conversions to other media types.
"Work" shall mean the work of authorship, whether in Source or
Object form, made available under the License, as indicated by a
copyright notice that is included in or attached to the work
(an example is provided in the Appendix below).
"Derivative Works" shall mean any work, whether in Source or Object
form, that is based on (or derived from) the Work and for which the
editorial revisions, annotations, elaborations, or other modifications
represent, as a whole, an original work of authorship. For the purposes
of this License, Derivative Works shall not include works that remain
separable from, or merely link (or bind by name) to the interfaces of,
the Work and Derivative Works thereof.
"Contribution" shall mean any work of authorship, including
the original version of the Work and any modifications or additions
to that Work or Derivative Works thereof, that is intentionally
submitted to Licensor for inclusion in the Work by the copyright owner
or by an individual or Legal Entity authorized to submit on behalf of
the copyright owner. For the purposes of this definition, "submitted"
means any form of electronic, verbal, or written communication sent
to the Licensor or its representatives, including but not limited to
communication on electronic mailing lists, source code control systems,
and issue tracking systems that are managed by, or on behalf of, the
Licensor for the purpose of discussing and improving the Work, but
excluding communication that is conspicuously marked or otherwise
designated in writing by the copyright owner as "Not a Contribution."
"Contributor" shall mean Licensor and any individual or Legal Entity
on behalf of whom a Contribution has been received by Licensor and
subsequently incorporated within the Work.
2. Grant of Copyright License. Subject to the terms and conditions of
this License, each Contributor hereby grants to You a perpetual,
worldwide, non-exclusive, no-charge, royalty-free, irrevocable
copyright license to reproduce, prepare Derivative Works of,
publicly display, publicly perform, sublicense, and distribute the
Work and such Derivative Works in Source or Object form.
3. Grant of Patent License. Subject to the terms and conditions of
this License, each Contributor hereby grants to You a perpetual,
worldwide, non-exclusive, no-charge, royalty-free, irrevocable
(except as stated in this section) patent license to make, have made,
use, offer to sell, sell, import, and otherwise transfer the Work,
where such license applies only to those patent claims licensable
by such Contributor that are necessarily infringed by their
Contribution(s) alone or by combination of their Contribution(s)
with the Work to which such Contribution(s) was submitted. If You
institute patent litigation against any entity (including a
cross-claim or counterclaim in a lawsuit) alleging that the Work
or a Contribution incorporated within the Work constitutes direct
or contributory patent infringement, then any patent licenses
granted to You under this License for that Work shall terminate
as of the date such litigation is filed.
4. Redistribution. You may reproduce and distribute copies of the
Work or Derivative Works thereof in any medium, with or without
modifications, and in Source or Object form, provided that You
meet the following conditions:
(a) You must give any other recipients of the Work or
Derivative Works a copy of this License; and
(b) You must cause any modified files to carry prominent notices
stating that You changed the files; and
(c) You must retain, in the Source form of any Derivative Works
that You distribute, all copyright, patent, trademark, and
attribution notices from the Source form of the Work,
excluding those notices that do not pertain to any part of
the Derivative Works; and
(d) If the Work includes a "NOTICE" text file as part of its
distribution, then any Derivative Works that You distribute must
include a readable copy of the attribution notices contained
within such NOTICE file, excluding those notices that do not
pertain to any part of the Derivative Works, in at least one
of the following places: within a NOTICE text file distributed
as part of the Derivative Works; within the Source form or
documentation, if provided along with the Derivative Works; or,
within a display generated by the Derivative Works, if and
wherever such third-party notices normally appear. The contents
of the NOTICE file are for informational purposes only and
do not modify the License. You may add Your own attribution
notices within Derivative Works that You distribute, alongside
or as an addendum to the NOTICE text from the Work, provided
that such additional attribution notices cannot be construed
as modifying the License.
You may add Your own copyright statement to Your modifications and
may provide additional or different license terms and conditions
for use, reproduction, or distribution of Your modifications, or
for any such Derivative Works as a whole, provided Your use,
reproduction, and distribution of the Work otherwise complies with
the conditions stated in this License.
5. Submission of Contributions. Unless You explicitly state otherwise,
any Contribution intentionally submitted for inclusion in the Work
by You to the Licensor shall be under the terms and conditions of
this License, without any additional terms or conditions.
Notwithstanding the above, nothing herein shall supersede or modify
the terms of any separate license agreement you may have executed
with Licensor regarding such Contributions.
6. Trademarks. This License does not grant permission to use the trade
names, trademarks, service marks, or product names of the Licensor,
except as required for reasonable and customary use in describing the
origin of the Work and reproducing the content of the NOTICE file.
7. Disclaimer of Warranty. Unless required by applicable law or
agreed to in writing, Licensor provides the Work (and each
Contributor provides its Contributions) on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
implied, including, without limitation, any warranties or conditions
of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A
PARTICULAR PURPOSE. You are solely responsible for determining the
appropriateness of using or redistributing the Work and assume any
risks associated with Your exercise of permissions under this License.
8. Limitation of Liability. In no event and under no legal theory,
whether in tort (including negligence), contract, or otherwise,
unless required by applicable law (such as deliberate and grossly
negligent acts) or agreed to in writing, shall any Contributor be
liable to You for damages, including any direct, indirect, special,
incidental, or consequential damages of any character arising as a
result of this License or out of the use or inability to use the
Work (including but not limited to damages for loss of goodwill,
work stoppage, computer failure or malfunction, or any and all
other commercial damages or losses), even if such Contributor
has been advised of the possibility of such damages.
9. Accepting Warranty or Additional Liability. While redistributing
the Work or Derivative Works thereof, You may choose to offer,
and charge a fee for, acceptance of support, warranty, indemnity,
or other liability obligations and/or rights consistent with this
License. However, in accepting such obligations, You may act only
on Your own behalf and on Your sole responsibility, not on behalf
of any other Contributor, and only if You agree to indemnify,
defend, and hold each Contributor harmless for any liability
incurred by, or claims asserted against, such Contributor by reason
of your accepting any such warranty or additional liability.
END OF TERMS AND CONDITIONS
APPENDIX: How to apply the Apache License to your work.
To apply the Apache License to your work, attach the following
boilerplate notice, with the fields enclosed by brackets "[]"
replaced with your own identifying information. (Don't include
the brackets!) The text should be enclosed in the appropriate
comment syntax for the file format. We also recommend that a
file or class name and description of purpose be included on the
same "printed page" as the copyright notice for easier
identification within third-party archives.
Copyright [yyyy] [name of copyright owner]
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
+30
View File
@@ -0,0 +1,30 @@
# Third-party components and attribution
Copyright 2026 Universal Humanoid contributors.
Original Universal Humanoid code and documentation are licensed under the Apache License,
Version 2.0; see [LICENSE](LICENSE). This notice is informational and does not modify that license.
Bundled third-party components retain their own copyright notices and license terms below;
the project license does not replace them.
| Included material | Attribution / license text |
|---|---|
| `rl/mjlab/` | mjlab developers, Apache-2.0; `rl/mjlab/LICENSE` |
| mjlab Isaac Lab-derived helpers | Isaac Lab Project Developers, BSD-3-Clause; `rl/mjlab/licenses/IsaacLab.txt` |
| mjlab Unitree G1/Go1 assets | Unitree Robotics / MuJoCo Menagerie, BSD-3-Clause; `rl/mjlab/licenses/mujoco_menagerie-unitree_g1.txt` and `mujoco_menagerie-unitree_go1.txt` |
| mjlab i2rt YAM assets | i2rt robotics / MuJoCo Menagerie, MIT; `rl/mjlab/licenses/mujoco_menagerie-i2rt_yam.txt`; additional d405 mesh is supplied with mjlab |
| `rl/rsl_rl/` | ETH Zurich, NVIDIA Corporation & Affiliates, BSD-3-Clause; `rl/rsl_rl/LICENSE` |
| RSL-RL pfrl-derived normalization | Preferred Networks, MIT; `rl/rsl_rl/licenses/pfrl.txt` |
| T1 XML and referenced meshes | BoosterRobotics, BSD-3-Clause; `rl/mjlab/asset_zoo/robots/booster_t1/assets/LICENSE` |
Release metadata, source commits, retained modifications and inventories are described in
[docs/VENDOR.md](docs/VENDOR.md). Changed mjlab runtime files carry `[uh-vendor]` notices;
RSL-RL algorithm source is unchanged. License texts are included in built package resources.
The modular workflow, vendored-package layout and stage-specific operating-manual design draw
inspiration from Jumper. No Jumper hexapod controller, native-backend fork, credentials or .env
content is included. Ancillary third-party license texts were retained from that reference
project for the corresponding embedded components.
External asset overrides and any additional models retain their own license obligations.
ONNX research bundles contain no model/mesh redistribution and do not imply hardware approval.
+193
View File
@@ -0,0 +1,193 @@
# Universal Humanoid
**An agent-ready framework for modular humanoid reinforcement learning.**
Universal Humanoid connects AI coding assistants with a reproducible robotics workflow: define tasks, iterate on rewards, train policies, evaluate behavior, replay results, and export inference bundles. Built on **mjlab**, **MuJoCo Warp**, and **RSL-RL**, it separates robot-specific configuration from shared training and evaluation tools.
“Universal” describes an extensible engineering workflow—not a single policy that controls every robot without adaptation.
[中文](README.zh.md) · [Setup](docs/SETUP.md) · [Robot checks](docs/NEW_ROBOT.md) · [Training operations](docs/TRAINING_OPERATIONS.md) · [Architecture](docs/ARCHITECTURE.md) · [Agent manuals](AGENTS.md) · [CLI](docs/CLI.md) · [Validation](docs/VALIDATION.md)
## Overview
```text
Developer / AI coding assistant
│
Task & reward configuration
│
Robot-specific adapter
│
mjlab → MuJoCo Warp → RSL-RL PPO
│
Checkpoint & experiment record
│
Evaluation & visual replay
│
Verified ONNX research bundle
```
Use the CLI directly or let a coding assistant operate the same commands. The framework does not require a particular model provider or a separate hosted agent service; assistants use their own configured models and tools.
## Features
- **Vendored runtime:** mjlab and RSL-RL copies under `rl/`, bundled G1/T1 assets, editable installation, hashed dependency lock and source inventory.
- **Robot conformance:** one-command checks for foot sites, IMU frames, actual contact response, PD/limits, joint order and reset posture.
- **Modular tasks:** lazy-loaded robot/environment factories behind a shared CLI.
- **Reward iteration:** strict experiment JSON, explicit weight/parameter overrides, and registered Python reward functions.
- **GPU training:** batched MuJoCo Warp simulation with RSL-RL PPO, committed intermediate checkpoints and optimizer continuation.
- **Training operations:** optional Linux background jobs, queryable iteration progress, guarded stop, and TensorBoard scalar curves in JSON/CSV/SVG.
- **Experiment provenance:** resolved configurations, source snapshots, dependency/asset fingerprints, and checkpoint hashes.
- **Measured evaluation:** checkpoint × evaluation-seed × four-command matrices, per-world randomized resets, noisy/push stress tests, mean/sample variance and explicitly defined failure rates.
- **Checkpoint selection:** explicit-metric ranking and paired matrix comparison; no automatic deployment promotion.
- **Replay:** MP4, per-step joint/torque CSV before reset, and mjlab's native/Viser viewer integrations.
- **Policy export:** ONNX with observation/action contracts, resolved joint order, numerical parity checks, and integrity manifests.
- **Agent guidance:** setup-env, new-task, train, controls, deploy and cross-stage RL manuals, with environment scripts and layout/documentation synchronization tests.
## Project status
**Early-stage research framework.** Backends, robot adapters and example model resources are now included. The reference installation targets Linux x86-64 with NVIDIA CUDA; portability does not imply every platform/GPU or arbitrary robot is supported.
| Robot | Task | Integration status |
|---|---|---|
| Unitree G1, 29 DoF | Flat-ground velocity tracking | Adapter wired to mjlab's adapted G1 model; bounded pipeline validation recorded |
| Booster T1, 23 DoF | Flat-ground velocity tracking | Included adapter and BSD-licensed model/meshes; bounded validation recorded |
| Other humanoids | — | Add and validate a robot/task adapter; asset discovery alone does not imply support |
**Backend policy:** install this project, not separate `mjlab`/`rsl-rl-lib` distributions. The retained backend changes—including simulation warning handling and T1 velocity-command support—are recorded in [VENDOR.md](docs/VENDOR.md), not hidden in an external working tree. See [Setup](docs/SETUP.md) and the [release checklist](docs/RELEASING.md).
The vendored build has passed [relocated installation, robot-gate and wheel checks](docs/PORTABILITY_VALIDATION.md) on the reference GPU host.
The included experiments are starting recipes, not pretrained walking policies. Initial validation used short training runs whose policies still fell. G1 and T1 use separate policies. See [recorded evidence and limitations](docs/VALIDATION.md).
## Getting started
### Explore without simulation dependencies
From the repository root, with Python 3.11 or newer:
```bash
PYTHONPATH=src python -m universal_humanoid list
PYTHONPATH=src python -m universal_humanoid plan --config experiments/g1_smoke.json
```
### Configure a simulation environment
Training requires a supported NVIDIA GPU/driver. Install the locked environment; no external backend or asset path is needed for G1/T1:
```bash
python3 .agents/skills/setup-env/scripts/detect.py
uv sync --locked --extra dev --python 3.13
./scripts/uh doctor
./scripts/uh verify-vendor
./scripts/uh check-robot --config experiments/g1_smoke.json --output runs/g1-check.json
./scripts/uh smoke --config experiments/g1_smoke.json --steps 16
```
Standard `pip install -e '.[dev]'` is also supported; [Setup](docs/SETUP.md) explains the hashed pip lock alternative. `doctor` checks installation/CUDA, while `check-robot` checks the actual adapter and measured contacts. Use the setup-env scripts for a logged multi-gate report. Legacy external `MJLAB_ROOT` overrides are rejected.
### Run a bounded end-to-end experiment
```bash
# Two PPO iterations, evaluation, then a research-only ONNX bundle.
# Use a NEW output directory for every run.
./scripts/uh pipeline --config experiments/g1_smoke.json \
--output runs/g1-first --eval-steps 250 --eval-envs 4
./scripts/uh verify-bundle runs/g1-first/bundle --inference
```
This validates the pipeline, not locomotion quality. Select `experiments/t1_smoke.json` for the included T1 adapter and assets.
## Reward → train → evaluate → replay → export
```bash
./scripts/uh rewards --config experiments/g1_flat.json
./scripts/uh plan --config experiments/g1_reward_candidate.json
# Longer training: choose a resource budget before launching.
./scripts/uh train --config experiments/g1_flat.json --output runs/g1-baseline
./scripts/uh evaluate --checkpoint runs/g1-baseline/final.pt \
--output runs/g1-baseline/evaluation.json --steps 1500 --num-envs 8 \
--command 0.3 0 0 --video runs/g1-baseline/replay.mp4
./scripts/uh play --checkpoint runs/g1-baseline/final.pt --viewer viser
# Desktop viewer: --viewer native (requires a display).
./scripts/uh export --checkpoint runs/g1-baseline/final.pt \
--evaluation runs/g1-baseline/evaluation.json --output bundles/g1-baseline
./scripts/uh verify-bundle bundles/g1-baseline --inference
```
Change tuning in a new experiment JSON. Add reward mechanisms in `src/universal_humanoid/rewards.py` and register them in `FUNCTIONS`. Unknown term names, parameters, and function references are rejected.
Evaluate/export a baseline **before** editing source. Replay and export check source, asset, and dependency identity; historical sources can be restored from a trusted run archive into a separate environment. Never overwrite shared dependencies to replay an old experiment.
Automation should read `run.json`, `pipeline.json`, and `evaluation.json`. Console output also contains backend diagnostics and is not a single JSON document.
## Monitor learning and manage experiments
```bash
./scripts/uh train --config experiments/g1_flat.json --output runs/g1-long --background
./scripts/uh status runs/g1-long
./scripts/uh curves runs/g1-long --output analysis/g1-curves
./scripts/uh checkpoints runs/g1-long
# Resume an explicitly selected checkpoint into a NEW run:
./scripts/uh train --resume runs/g1-long/model_0.pt --iterations 200 \
--output runs/g1-continued --background
./scripts/uh benchmark --checkpoints runs/g1-long/final.pt \
--seeds 101 202 303 --profiles randomized stress \
--steps 1000 --num-envs 8 --output analysis/g1-matrix
./scripts/uh play --checkpoint runs/g1-long/final.pt \
--measure analysis/g1-joints.csv --steps 1000 --profile stress
```
These are separate lifecycle examples, not a script to launch all jobs simultaneously. Wait for
checkpoints to be committed and reserve GPU capacity before evaluating beside a running trainer.
Matrix seeds control evaluation randomness, not independent training replicas. Read tracking,
action-noise and episode-length curves—not just elapsed time. See [Training operations](docs/TRAINING_OPERATIONS.md)
for stopping, continuation semantics, statistics, checkpoint selection and CSV timing.
## Working with a coding assistant
Start with [AGENTS.md](AGENTS.md) and the [RL engineer playbook](.agents/skills/rl-engineer/SKILL.md). For example:
> Establish a G1 baseline, propose a reward revision, and test both within an agreed training budget. Compare tracking error and failure events under the same evaluation protocol, record a replay, and export a research bundle. Do not connect to physical hardware.
Reward totals alone are not a valid comparison across different reward definitions. Use task metrics, visual inspection, multiple seeds, and appropriate evaluation horizons.
## Assets and extensions
```bash
# Optional inventory of additional models, not needed for the bundled G1/T1 tasks:
export UNITREE_ROBOTS_ROOT=/path/to/unitree_ros/robots
./scripts/uh assets
./scripts/uh inspect-asset "$UNITREE_ROBOTS_ROOT/g1_description/g1_29dof.xml"
```
G1 training uses the bundled sensor/collision-adapted asset. T1 uses its bundled `T1_23dof.xml` and meshes; `BOOSTER_ASSETS_ROOT` is an optional explicit replacement, not a requirement. Raw robot descriptions are not silently substituted into a trained task. New morphologies need validated sensors, contacts, control gains, joint ordering, and observation/action definitions. See [extension seams](docs/ARCHITECTURE.md).
## Development
Git, license, dependency-lock and CI details are recorded in the [repository baseline](docs/REPOSITORY_BASELINE.md).
See [CONTRIBUTING.md](CONTRIBUTING.md) for change boundaries and required checks.
With the configured runtime and test dependencies installed:
```bash
./scripts/check
```
Core tests can run without a simulator; adapter tests require mjlab/torch. Unit tests and GPU validation are separate forms of evidence. Before publishing a release, follow [RELEASING.md](docs/RELEASING.md).
## Scope and safety
Not currently provided: physical robot transport, state estimation, watchdog/E-stop supervision, ROS/DDS mappings, automatic morphology conversion, a universal zero-shot policy, rough-terrain task coverage, motion imitation, multi-GPU training, or native-CPU RL.
Exported bundles are for **research/inference**, not robot executables. Every bundle has `hardware_approved: false`; no CLI command enables motors. Load only trusted PyTorch checkpoints—the upstream loader uses pickle. Hashes detect accidental changes; they are not signatures or a sandbox.
## Acknowledgments and licensing
The workflow design draws inspiration from Jumper's modular task and engineering conventions. Simulation and learning are provided by mjlab, MuJoCo Warp, and RSL-RL. See [NOTICE.md](NOTICE.md) for component and asset attribution.
Original project code and documentation are licensed under [Apache-2.0](LICENSE), following Jumper's licensing approach. Bundled dependencies and robot assets retain their original licenses; see [NOTICE.md](NOTICE.md).
+155
View File
@@ -0,0 +1,155 @@
# Universal Humanoid
**面向 AI 代码智能体的模块化人形机器人强化学习框架。**
Universal Humanoid 将代码智能体与可复现的机器人工程流程连接起来,支持任务定义、奖励迭代、策略训练、定量评估、可视化回放和推理策略包导出。框架基于 **mjlab、MuJoCo Warp 和 RSL-RL**,将机器人专属配置与共享训练工具分离。
“Universal”指的是**可扩展的工程架构**,而不是一个无需适配就能控制所有机器人的策略。
[English](README.md) · [环境配置](docs/SETUP.md) · [训练运维](docs/TRAINING_OPERATIONS.md) · [架构设计](docs/ARCHITECTURE.md) · [Agent 手册](AGENTS.md) · [验证记录](docs/VALIDATION.md)
## 工作流程
```text
开发者 / AI Coding Agent
↓
任务定义与奖励配置
↓
机器人适配 → mjlab → MuJoCo Warp → RSL-RL PPO
↓
检查点与实验记录
↓
定量评估 + 可视化回放
↓
经过校验的 ONNX 研究用策略包
```
可以直接使用 CLI,也可以让支持文件编辑与命令执行的代码智能体操作相同工具。项目不绑定某个模型厂商,不要求额外部署专用 Agent 服务;智能体使用各自配置的模型和工具。
## 核心能力
- **仓库内后端**:`rl/mjlab`、`rl/rsl_rl`、G1/T1 资产随项目安装,提供带哈希的依赖锁和源码清单。
- **机器人自动检查**:一次检查脚部 site、IMU 参考系、实际接触反馈、PD/限幅、关节顺序及初始姿态。
- **模块化任务**:机器人、环境和训练配置通过统一入口组合。
- **奖励迭代**:严格校验实验 JSON,支持权重、参数和 Python 奖励函数扩展。
- **GPU 强化学习**:MuJoCo Warp 批量仿真与 RSL-RL PPO 训练,支持中间检查点及优化器续训。
- **训练运维**:Linux 后台任务、进度查询、安全停止,以及 TensorBoard 曲线读取与 JSON/CSV/SVG 输出。
- **实验可追溯**:保存配置、源码快照、依赖与资产指纹、检查点哈希。
- **批量评估**:多个检查点 × 评估 seed × 站立/前进/横移/转向,汇总均值、样本方差与定义明确的失败指标。
- **压力评估**:各世界分别随机初始化,可叠加观测噪声和随机速度扰动,记录实际扰动次数。
- **检查点选择**:按显式指标排序,支持同协议矩阵的配对比较,不自动批准部署。
- **可视化回放**:MP4、重置前逐步关节/力矩 CSV,集成 mjlab 原生与 Viser 查看器。
- **策略导出**:ONNX、观测/动作契约、关节顺序、推理一致性和完整性校验。
- **Agent 手册体系**:setup-env、new-task、train、controls、deploy 及跨阶段 RL 手册;附环境脚本和文档/接口一致性测试。
## 项目状态
项目处于**早期研究阶段**。后端源码、适配器和示例模型已纳入仓库;参考锁定环境为 Linux x86-64 + NVIDIA CUDA,不代表支持所有平台、GPU 或任意机器人。
| 机器人 | 示例任务 | 接入状态 |
|---|---|---|
| Unitree G1,29 自由度 | 平地速度跟踪 | 使用 mjlab 已适配模型,已记录短流程验证结果 |
| Booster T1,23 自由度 | 平地速度跟踪 | 已包含适配器及 BSD 许可模型/网格,已记录短流程验证结果 |
| 其他人形机器人 | 待扩展 | 需新增并验证适配器;发现模型文件不等于完成支持 |
**安装规则:**直接安装本项目,不要再单独安装 `mjlab` 或 `rsl-rl-lib`。包括 `sim.py`、速度命令与 T1 适配在内的修改均已纳入仓库,并在[后端来源说明](docs/VENDOR.md)中记录;不再依赖外部未提交的工作区。
实验配置是起点,不是预训练行走策略。初始短训练策略仍会跌倒;G1 和 T1 分别训练,不共享一个通用策略。具体证据与限制见 [VALIDATION.md](docs/VALIDATION.md)。
## 快速开始
### 无仿真依赖地查看任务
在仓库根目录,使用 Python 3.11 或更新版本:
```bash
PYTHONPATH=src python -m universal_humanoid list
PYTHONPATH=src python -m universal_humanoid plan --config experiments/g1_smoke.json
```
### 配置运行环境
训练需要兼容的 NVIDIA GPU/驱动。G1/T1 无需外部后端或资产路径,按锁文件安装即可:
```bash
python3 .agents/skills/setup-env/scripts/detect.py
uv sync --locked --extra dev --python 3.13
./scripts/uh doctor
./scripts/uh verify-vendor
./scripts/uh check-robot --config experiments/g1_smoke.json --output runs/g1-check.json
./scripts/uh smoke --config experiments/g1_smoke.json --steps 16
```
也支持标准 `pip install -e '.[dev]'`;[环境配置](docs/SETUP.md)提供带哈希的 pip 锁定安装方式。`doctor` 检查环境,`check-robot` 检查适配器和实际接触;setup-env 脚本可统一执行并保存各项日志。旧的 `MJLAB_ROOT` 外部后端覆盖已被拒绝。
### 运行一个小规模完整流程
```bash
./scripts/uh pipeline --config experiments/g1_smoke.json \
--output runs/g1-first --eval-steps 250 --eval-envs 4
./scripts/uh verify-bundle runs/g1-first/bundle --inference
```
每次使用新的输出目录。`*_smoke.json` 仅训练两轮,用于验证工程流程,**不用于行走能力验收**。使用已包含的 T1 适配器时,改用 `experiments/t1_smoke.json`。完整训练、评估、视频和导出命令见 [English README](README.md)。
## 看训练过程,而不只看耗时
```bash
./scripts/uh train --config experiments/g1_flat.json --output runs/g1-long --background
./scripts/uh status runs/g1-long
./scripts/uh curves runs/g1-long --output analysis/g1-curves
./scripts/uh checkpoints runs/g1-long
./scripts/uh stop runs/g1-long
# 从已有的、已提交的检查点再训练 200 轮;始终使用新目录。
./scripts/uh train --resume runs/g1-long/model_0.pt --iterations 200 \
--output runs/g1-continued --background
```
这些是生命周期操作示例,不应无等待地全部连跑。先确认检查点已提交;训练与评估共用 GPU 时需预留显存。`curves` 同时读取跟踪项、动作分布标准差和回合长度,缺失曲线会明确报告。
`benchmark` 默认的 seed 是**评估 seed**;如需研究训练种子差异,应分别训练不同 seed 的配置。默认评估采用随机初始状态,`--profile stress` 额外开启任务观测噪声与随机速度扰动。续训恢复优化器等训练状态,但不是中断轨迹的逐位复现。
完整的矩阵评估、检查点选择、失败率口径与 `play --measure` 示例见[训练运维手册](docs/TRAINING_OPERATIONS.md)。
## 让代码智能体参与实验
可向代码智能体提出:
> 按 AGENTS.md 和 rl-engineer 手册,为 G1 建立 baseline 并提出奖励改进。在约定预算内训练,用相同评估协议比较跟踪误差和失败事件,记录视频并导出研究用策略包,不连接真机。
规范入口:[AGENTS.md](AGENTS.md)、[RL 工程手册](.agents/skills/rl-engineer/SKILL.md)。不同奖励定义下的总奖励不能直接比较;需要结合任务指标、视频、多种随机种子和足够长的评估时间。
## 项目结构
| 位置 | 用途 |
|---|---|
| `rl/mjlab/`、`rl/rsl_rl/` | 仓库内固定后端源码与许可证 |
| `uv.lock` | 完整依赖解析与制品哈希 |
| `src/universal_humanoid/robot_specs/` | 独立声明的机器人检查契约 |
| `src/universal_humanoid/registry.py` | 机器人/任务注册与工厂引用 |
| `src/universal_humanoid/tasks.py` | 配置适配、奖励注入、评估协议 |
| `src/universal_humanoid/rewards.py` | 可扩展的奖励机制 |
| `experiments/` | 独立实验配方 |
| `src/universal_humanoid/runtime.py` | 训练、评估、视频与交互回放 |
| `src/universal_humanoid/deployment.py` | 策略导出、契约校验与纯推理消费者 |
| `docs/` | 环境、架构、验证和发布说明 |
| `runs/`、`bundles/` | 运行生成的产物,不纳入版本控制 |
迁移安装、独立环境、机器人门禁、wheel 资源及完整短流程的实测结果见[可移植性验证](docs/PORTABILITY_VALIDATION.md)。
新增机器人需要验证模型、传感器、碰撞、控制参数及关节顺序。原始机器人描述不会被自动替换进训练任务;详细扩展方式见[架构文档](docs/ARCHITECTURE.md)。
## 边界与安全
当前不包含真机通信、状态估计、安全监督器、急停、ROS/DDS 映射、自动形态转换、复杂地形任务、动作模仿、多 GPU 训练或原生 CPU 强化学习后端。
导出产物是**研究/推理用策略包**,不是机器人可执行程序。所有包均标记 `hardware_approved: false`,没有启用电机的命令。仅加载可信检查点;文件哈希不是数字签名或安全沙箱。
## 致谢与发布
Git、许可证、依赖锁与 CI 的配置和本地检查结果见[仓库基线](docs/REPOSITORY_BASELINE.md)。
工程流程设计参考 Jumper 的模块化任务与操作规范,仿真和学习依托 mjlab、MuJoCo Warp 与 RSL-RL。组件和资产归属见 [NOTICE.md](NOTICE.md)。
参考 Jumper,本项目原创代码和文档采用 [Apache-2.0](LICENSE);内置依赖和机器人资产保留原许可证,见 [NOTICE.md](NOTICE.md)。对外发布前仍需完成[发布检查清单](docs/RELEASING.md)中的跨机器验证和来源审核。
+102
View File
@@ -0,0 +1,102 @@
# Architecture and extension seams
## Layers
```text
Coding agent / operator
AGENTS.md + rl-engineer playbook
|
Unified CLI (no robot-specific branches)
strict experiment JSON, bounded commands, durable run/pipeline status
|
Metadata registry → lazy task factory → config adapter
robot facts mjlab baseline reward overrides / evaluation protocol
|
mjlab ManagerBasedRlEnv
observation | action | command | reward | termination | event | curriculum
|
MuJoCo Warp (CUDA) → RSL-RL PPO
|
checkpoint + source/asset identity
|
fixed-command evaluation + video → ONNX parity → research bundle
```
## Relationship to Jumper
Reuse the architectural principles: independent task modules, centralized commands, explicit
agent playbooks, and a deployment contract resolved from the actual environment. Do not copy
its hexapod control assumptions, native backend patches, RKNN conversion or Rust FSM. This
project uses the existing mjlab `Simulation` and manager API without monkey-patching it. Native
MuJoCo is used only to inspect MJCF and by the optional viewer, not as a hidden RL fallback.
## Backend and adapter contract
Registry entries reference environment, training and asset factories through importable Python
paths. A compatible backend must provide those interfaces; the registry does not download or
convert missing adapters. Fingerprints include **all Python files**, not merely git HEAD.
The validated mjlab 1.6.0 source, its two inherited runtime patches, T1 extension and RSL-RL 5.5.1
are now vendored under `rl/`, installed as top-level packages with the project. Runtime guards
reject an external shadowing backend. `uv.lock` fixes transitive dependencies; `rl/VENDOR.json`
checks source and asset inventory. See [Vendor provenance](VENDOR.md) and [Setup](SETUP.md).
G1 and T1 use bundled adapted models with original asset licenses. `BOOSTER_ASSETS_ROOT` is an
optional explicit T1 override, not a hidden HOME default. Raw Unitree ROS XML can be inspected
but is not mechanically or semantically interchangeable with those models. `check-robot` checks
resolved interfaces against independent declarations in `robot_specs/`, then measures short
contact/PD response. See [Robot checks](NEW_ROBOT.md).
## Add a robot/task without changing the CLI
1. Provide a robot `EntityCfg` factory: MJCF `spec_fn`, PD/motor definitions, initial posture,
collision rules, foot sites and required sensors. Do not infer motor ratings from joint names.
2. Provide an environment factory `factory(play=False)` returning an independent config and an
RSL-RL config factory. The initial contract is a flat velocity task: entity `robot`, command
`twist`, `joint_pos` position action, actor/critic observation groups, reset events.
3. Add metadata in `registry.py`: ID, robot, expected action DoF, lazy factory paths, asset
reference (Path or callable returning Path), and number of parent levels covering its assets.
4. Add an experiment. No train/evaluate/export command needs robot-specific changes.
5. Validate all task seams with the playbook. The current exporter deliberately requires one
built-in position actuator per controlled joint. Tendons, coupled drives, recurrent networks,
multiple action groups and history stacks need an explicitly versioned contract extension.
Task registry extensibility is not automatic adaptation to arbitrary morphology. A future
tracking/manipulation task needs a separate task protocol rather than abusing the twist interface.
## Reproducibility and evaluation
Runs contain the input experiment, resolved environment/agent configuration, checkpoint hashes,
Python/config source archive, both backend source identities, dependency lock/versions and
source/asset fingerprints. Bundled and optional external model assets retain their original
licenses; meshes are not copied into training source archives. Replay/export refuse changed source/assets/runtime
versions. Archives are for manual trusted restoration, never automatic execution.
Comparison intentionally permits different reward/source hashes but requires the same task and
evaluation protocol. Inspect provenance to ensure differences really are reward-only; the CLI does
not prove scientific causal isolation. Use multiple seeds, more than one command and adequate
horizons. `status=completed` means the operation finished, not the policy passed a quality gate.
## Resource and trust model
No remote shell, multi-agent scheduler, cloud logger or motor transport is started. Training is
bounded by the experiment's iterations. `train --background` optionally launches a detached local
Linux worker with progress, logs and PID-guarded stop; foreground execution remains the default.
`pipeline` records the current stage and failure. Graceful interruption records `interrupted`;
SIGKILL/power loss can leave stale status, which `status` reports as lost when the recorded process
is dead. Never rewrite a failed/interrupted run to completed.
`training.py` atomically publishes intermediate checkpoints, records completed-update counts and
observes the backend logger without editing external packages. `jobs.py` / `worker.py` manage
local process lifecycle; `curves.py` reads scalar logs without a server. `evaluation.py` defines
versioned randomized/stress profiles; `benchmark.py` aggregates seed-level reports and selects
explicit candidates. `measurement.py` records terminal joint states before manual reset.
Evaluation/resume may explicitly load a committed checkpoint from an incomplete run after
checking its hash and provenance. Export still requires a completed run. Resume uses a new output
branch and restores optimizer/model/curriculum-counter state, not simulator or RNG trajectories.
See [Training operations](TRAINING_OPERATIONS.md) for the statistical and lifecycle contracts.
Hash manifests are integrity evidence, not cryptographic signatures. Checkpoints and Python task
factories are executable trusted inputs. Source paths and git status are recorded; `.env` and
credentials are not copied. Binary policies cannot establish physical safety.
+46
View File
@@ -0,0 +1,46 @@
# CLI reference
Run `./scripts/uh COMMAND --help` or the installed `uh COMMAND --help`. All paths are explicit;
run/output directories must be new where the command creates an artifact. This table lists every
public command and its named options (excluding automatic `--help`). Layout tests compare it to
the actual argparse parser; change the table with the interface.
| Command | Named options | Purpose / positional inputs |
|---|---|---|
| `list` | — | Registered task metadata |
| `doctor` | — | Installed backend, locked core versions and CUDA |
| `verify-vendor` | — | Checked-in source/asset inventory |
| `check-robot` | `--config --output --steps --spec` | Robot interface/mechanics/contact gate |
| `assets` | — | Discover explicitly configured optional external assets |
| `inspect-asset` | — | Native MJCF compile; positional model path |
| `plan` | `--config` | Resolve experiment metadata without simulation |
| `rewards` | `--config` | Resolve reward functions/weights/parameters |
| `smoke` | `--config --steps --num-envs` | Bounded zero-action GPU probe |
| `train` | `--config --resume --iterations --background --output` | PPO or additional updates from a committed checkpoint |
| `pipeline` | `--config --output --eval-steps --eval-envs` | Bounded train/evaluate/export sequence |
| `evaluate` | `--checkpoint --output --steps --num-envs --command --video --seed --profile --measure` | Fixed-command evaluation with optional measurement/video |
| `play` | `--checkpoint --viewer --measure --steps --seed --profile --command --video` | Interactive viewer, or bounded headless measured replay |
| `status` | — | Positional run directory; progress and process state |
| `stop` | — | Positional background run directory; guarded graceful stop |
| `checkpoints` | — | Positional run directory; committed checkpoint inventory |
| `curves` | `--output --match --window` | Positional run directory; TensorBoard JSON/CSV/SVG |
| `benchmark` | `--checkpoints --output --seeds --profiles --steps --num-envs` | Checkpoint × evaluation-seed × command × profile matrix |
| `select-checkpoint` | `--output --metric` | Positional completed matrix summary; explicit-metric ranking |
| `export` | `--checkpoint --evaluation --output` | Research-only ONNX bundle |
| `verify-bundle` | `--inference` | Positional bundle; integrity and optional inference parity |
| `compare` | — | Positional baseline/candidate reports or single-checkpoint matrix summaries |
## Important semantics
- `doctor`, `verify-vendor` and `check-robot` failures give a nonzero exit. A failed robot gate
writes its JSON report; a model construction failure cannot provide downstream dynamic checks.
- `--iterations` on resume is additional updates; it is not the target total update index.
- `benchmark --seeds` sets evaluation randomness, not independent training runs.
- `play --measure` selects bounded headless replay. Its seed/profile/command/steps/video options
describe that mode; ordinary `play --viewer` is the interactive viewer path.
- Default evaluation profile is randomized; stress adds configured observation noise and velocity
kicks. Nominal fixed replicas do not become independent samples merely by increasing worlds.
- Backend diagnostics may accompany JSON on stdout. Read generated JSON artifacts for automation.
References: [Setup](SETUP.md), [Robot checks](NEW_ROBOT.md), [Training operations](TRAINING_OPERATIONS.md),
[Bundle contract](DEPLOYMENT.md). Hardware approval is not a CLI operation.
+61
View File
@@ -0,0 +1,61 @@
# Research bundle schema v1
An export is an immutable directory:
```text
policy.onnx deterministic actor, with trained observation normalization
contract.json observation/action ordering, affine transforms, simulation PD settings
reference.json one raw observation and its expected unclipped policy action
evaluation.json fixed-protocol rollout evidence bound to the checkpoint/source hashes
experiment.json original experiment
agent.json runner architecture/configuration
provenance.json source/assets/runtime versions and hashes
manifest.json SHA-256 of every payload file, checkpoint hash, ONNX parity results
```
The bundle contains no robot meshes, checkpoint pickle, board executable or motor transport.
It is portable for **CPU ONNX inference**, not a self-contained simulation environment. Replay
still needs the recorded task code, asset checkout and compatible mjlab environment.
## Consumer contract
`PolicyRuntime(Path(bundle)).predict(observation)` accepts a finite float32 `[1, observation_size]`
array. Observation terms must be concatenated in the exact `start:stop` order. Read each term's
function identity in `contract.json`; implementation semantics are in the recorded sources.
For current flat tasks, terms are base linear/angular velocity from the configured IMU sensors,
projected gravity, joint position relative to default, joint velocity relative to default,
previous clipped raw action, and `(vx, vy, yaw_rate)` commands. Sensor frame/site and filtering
matter; a base estimator cannot be substituted without validating parity. Joint observation
ordering is explicitly separate from action ordering. Units are SI; angles in radians.
Normalization is inside ONNX. Do **not** normalize twice. On reset, previous action is zero.
During inference:
```text
raw_action = policy(observation)
clipped_action = clip(raw_action, -raw_clip, raw_clip)
joint_target_rad = offset + scale * clipped_action
```
The consumer returns all three arrays and does not send them anywhere. Host PD/control frequency
must match the recorded policy/physics timestep semantics. Simulation gain and torque data are
metadata, not a certified motor configuration. Joint-limit clamping is not silently added to
inference because it would change training semantics; a physical safety controller must be
separately designed, validated and may refuse the policy.
## Export checks and limits
Export requires a completed local run and matching evaluation report. It tests ONNX checker
validity and action parity at 16 actual rollout observations (`atol=1e-5`, `rtol=1e-4`), then
writes file hashes. `verify-bundle --inference` checks integrity and one saved reference.
This detects common swapped-joint/normalizer/version mistakes, but is not a complete behavioral
proof. The current exporter requires MLP, one joint-position action group, no per-joint action
clip, one position actuator per joint and batch size one. It does not generate a robot driver.
A report can show poor tracking and many falls and still be packaged for research. All bundles
are `hardware_approved: false`. There is no release-quality gate or hardware approval command.
Before physical deployment, implement and validate bus index mapping, estimator/calibration,
IMU frame, latency, torque/position bounds, thermal limits, watchdog, controlled startup/shutdown,
E-stop and supervised sim-to-real tests. Require explicit operator authorization for actual motion.
+72
View File
@@ -0,0 +1,72 @@
# Adding and checking a robot/task
Asset compilation is only the first step. A robot adapter must explicitly define model mechanics,
sensors, controls, observations, reset posture and task semantics. There is no automatic
conversion of an arbitrary MJCF into a validated walking task.
## One-command conformance check
```bash
./scripts/uh check-robot --config experiments/g1_smoke.json --output runs/g1-check.json
./scripts/uh check-robot --config experiments/t1_smoke.json --output runs/t1-check.json
```
The command builds the **actual nominal task**, then performs a bounded zero-action PD probe
(default 32 control steps, one CUDA world). No policy or hardware is loaded. It produces JSON
and exits nonzero if any check fails. The report contains these stable check IDs:
| Check ID | What is actually checked |
|---|---|
| `free_root` | Exactly one floating root joint |
| `foot_sites` | Named foot sites exist on declared bodies, with declared local positions/orientations and plausible initial height above flat ground |
| `imu` | Declared sensor names, types and dimensions; exact reference site, parent body, local position and orientation |
| `joint_order` | Resolved action joint order and observation joint order match the separately declared lists; no duplicate action names |
| `pd_and_limits` | One joint-transmission actuator per controlled hinge; enabled torque/joint limits, finite positive stiffness/nonnegative damping, expected gains, gear and ranges |
| `action_transform` | Resolved action scales/offsets and physics/policy timestep match the declaration |
| `initial_pose` | Default and actual root/joint pose, nominal +0.02 m reset height, zero initial velocities, hard-limit compliance and upright orientation |
| `initial_penetration` | Native MuJoCo contact distance at the initialized task pose stays within the declared penetration threshold |
| `contact_configuration` | Contact sensor resolves the declared foot bodies, filters terrain, requests found/force fields and has enabled collision geometry |
| `contact_response` | Each declared foot actually reports contact and positive force during the probe |
| `finite_probe` | Observations/rewards/joints/torques/contact forces stay finite, with no task termination during the bounded probe |
Foot force magnitude, initial local site placement, gains, limits, names and timing are retained
as evidence. A construction failure is reported rather than converted into a pass; checks that
require a successfully constructed environment cannot run after construction fails.
This checks interface consistency and a short contact/control response—not motor identification,
long-term balance, mesh realism, convergence, force accuracy or hardware safety. Initial penetration
uses native MuJoCo on the compiled model; runtime contacts use the actual Warp task. These are
complementary checks, not claims of bitwise physics parity.
## Adapter procedure
1. **Inspect assets and licensing.** Use `inspect-asset PATH`. Confirm mesh references, floating
root, units, mass/inertia, hinges, limits, actuator transmissions and model variant. Keep new
adapters in this project's package; do not silently edit a shared external repository.
2. **Define robot mechanics.** Implement an `EntityCfg` factory, explicit PD/effort settings,
reset pose, collisions, foot sites and IMU sensors. Record the source of hardware facts.
3. **Define the task.** Provide independent environment/PPO factories following
[ARCHITECTURE.md](ARCHITECTURE.md). Existing flat-velocity integration expects entity `robot`,
command `twist`, `joint_pos` actions, actor/critic groups, and base/joint reset events.
4. **Register metadata.** Add a `Task` in `src/universal_humanoid/registry.py` and a bounded
experiment in `experiments/`. The CLI itself should not gain robot-specific branches.
5. **Declare independent expectations.** Add `src/universal_humanoid/robot_specs/<robot>.json`,
using an existing declaration as a schema example. Review every joint order, gain, limit,
site, IMU frame and posture against intended mechanics. Do not auto-generate expected values
from the very environment being checked—that would validate a mistake against itself.
6. **Run the gate.** Use `check-robot --config ... --output NEW.json`. During development an
explicit `--spec FILE.json` can be used; it must match the task and schema. Unknown/missing
sites, changed PD gains, swapped order or absent contact must remain failures.
7. **Test behavior.** Add unit/negative tests, then bounded PPO, randomized/stress evaluation,
measured CSV/video replay and ONNX parity. A passing robot gate is not the final acceptance.
The supplied declarations encode the existing adapted G1 and T1 baselines; they are **simulation
contracts**, not certified hardware specifications. Updating an intended control setting requires
review of both the adapter and declaration, followed by new tests and run provenance.
## Nonstandard robots/tasks
This checker and exporter intentionally target flat-ground humanoids with one position actuator
per controlled hinge. Tendons, coupled actuators, prismatic joints, multiple action groups,
non-flat terrain or non-velocity tasks need an explicit extension of the contract and tests.
Do not relax a check until it passes while continuing to claim the old contract.
+102
View File
@@ -0,0 +1,102 @@
# Training-operations validation
Reference-environment checks on **2026-11-03**, using the Linux/RTX 5080 backend described in
[VALIDATION.md](VALIDATION.md). The files below are development evidence under ignored `runs/`,
not artifacts shipped with a fresh repository clone. Usage and statistical definitions are in
[TRAINING_OPERATIONS.md](TRAINING_OPERATIONS.md).
## Automated checks
**58 tests passed**; Ruff lint/format passed. One upstream TorchScript deprecation warning remains.
New checks cover seed-level sample variance and null statistics, duplicate/mismatched protocols,
paired comparisons, selection coverage, TensorBoard scalar reading and `/time` axis handling,
CSV joint ordering and projected torque, atomic checkpoint publication and update indexing,
committed-vs-uncommitted partial runs, Linux PID identity/stop, and both robots' evaluation profiles.
## Background training and real interruption/resume
- `runs/ops_g1_validation`: 16 environments × 24 steps × **4 updates**, launched with
`train --background`; state progressed to completed, with live progress and committed
`model_0.pt`, `model_3.pt`, `final.pt` visible through the CLI.
- `runs/ops_interrupt_validation`: a bounded 16-update job was deliberately stopped after its
first committed checkpoint. `status` observed it alive/training; `stop` sent SIGTERM;
worker and run both recorded **interrupted**, preserving `model_0.pt`.
- `runs/ops_resume_validation`: resumed that interrupted checkpoint for **2 additional updates**.
Final metadata reports **3 completed updates**, not 2 and not 4. Direct inspection of these
trusted generated checkpoint files confirmed Adam step counters **20 → 60** and the backend
common-step counter **24 → 72**. The optimizer was continued, not reinitialized.
No long convergence run, scheduler restart/reboot recovery, or multi-GPU orchestration was tested.
Process-loss/PID-reuse guards are covered by process-level tests, not a power-loss experiment.
## Actual matrix and checkpoint selection
`runs/ops_g1_matrix` contains **32 completed cells**:
- 2 checkpoints: after update 1 and update 4;
- 2 evaluation seeds: 101 and 202;
- 4 commands: standing, forward, lateral, turn;
- 2 profiles: randomized and stress;
- each cell: **2 worlds × 150 control steps**, dt 0.02 s, episode limit 5 s.
Every cell recorded **2 distinct initial states out of 2 worlds**. Stress cells recorded **71 total
velocity-kick events**. Repeating the same stress seed/command produced exactly matching initial
state arrays; XY RMSE differed by about `1.88e-5 m/s`, demonstrating why seeded draws are not a
promise of bitwise deterministic CUDA trajectories.
`table.csv`, `table.md`, and `summary.json` contain seed-level means/sample variances, failure
metrics and censoring. This 3 s observation horizon is intentionally short for engineering checks,
not sufficient policy acceptance evidence.
`runs/ops_g1_selection.json` ranked the earlier checkpoint first by equal-weight mean failure
incidence: **0.3333 vs 0.3750 events per simulated world-second**. This is not a general quality
ranking. Both policies fall; more updates did not automatically produce a better measured result.
`runs/ops_g1_baseline_matrix` and `runs/ops_g1_candidate_matrix` are single-checkpoint views copied
from the same 32 real cells, not additional independent experiments. `compare` successfully
produced paired-seed differences in `runs/ops_g1_comparison.json`.
## Training curves, not time spent
`runs/ops_g1_curves` contains JSON, CSV and parseable SVG extracted from actual TensorBoard events:
| Scalar | First | Last | Samples |
|---|---:|---:|---:|
| Linear tracking reward | 0.01648 | 0.05844 | 4 |
| Angular tracking reward | 0.004257 | 0.001208 | 4 |
| Mean action std | 1.00023 | 1.00013 | 4 |
| Mean episode length (control steps) | 12.0 | 51.65 | 4 |
Linear tracking/episode length increased, angular tracking decreased, and exploration std stayed
near 1. Four noisy points, random initial episode lengths and falling replays do **not** establish
learning or convergence. `/time` companion tags are no longer mislabeled as iteration curves.
## Measured replay and T1 coverage
| Evidence | Sampling | CSV rows | Failure events | Recorded kicks |
|---|---|---:|---:|---:|
| `runs/ops_g1_joints.csv` | 1 world × 150 steps × 29 joints | 4,350 | 1 | 2 |
| `runs/ops_t1_joints.csv` | 4 worlds × 150 steps × 23 joints | 13,800 | 6 | 7 |
Both used the stress profile and finite projected joint torques. CSV terminal-row counts matched
failure events × joint count (29 and 138), confirming terminal samples were retained before reset.
T1 recorded 4 distinct initial states out of 4 worlds. Its policy came from a separate two-update
run at `runs/ops_t1_validation`.
`runs/ops_g1_replay.mp4` decoded successfully: **150 frames, 50 FPS, 3 seconds**. A decoded frame
was visually inspected; robot and ground are visible. The matching measured report records a fall.
Interactive native/Viser viewing was not manually exercised in this revision.
A separate four-step evaluation, `runs/ops_g1_censored.json`, had no completed episodes:
completed-only failure fraction/return/duration were **null**, and both worlds were explicitly
censored rather than reported as successful episodes.
## Export regression and provenance
The updated evaluation report was accepted by ONNX export; `runs/ops_g1_bundle` passed
`verify-bundle --inference`. All outputs remain `hardware_approved: false`.
Earlier `ops_*_v1` development runs are retained. They predate the TensorBoard time-axis correction
and other final source changes and intentionally fail current source-identity checks. Historical
`g1_validation` / `t1_validation` runs also require their recorded sources for replay. Do not
rewrite provenance hashes to make old runs appear current.
+119
View File
@@ -0,0 +1,119 @@
# Vendoring, installation and robot-gate validation
Measured on **2026-11-03**, Linux x86-64 / RTX 5080. This revision replaces the external-backend
bootstrap with installed vendored packages. Evidence paths below are ignored development
artifacts, not files shipped in a fresh clone. Different hardware/platforms still need validation.
These measurements predate Git initialization and the subsequent license/CI-tool-lock changes.
Historical checkpoints retain their recorded lock and source identities; do not rewrite those
identities to replay them against a changed lock. Repository/CI checks are recorded separately in
[Repository baseline](REPOSITORY_BASELINE.md).
## Source and dependency boundary
- `rl/mjlab`: baseline commit `8bb2bc9e187797689ec82c78c35019afd2fd0376`, two captured runtime
patches and the included T1 adapter; details in [VENDOR.md](VENDOR.md).
- `rl/rsl_rl`: 5.5.1 Python files verified against installed release RECORD hashes before copying.
- Booster T1 asset source: commit `12a97516f57469078707afb587a5f84c1cb110d1`; only the selected
XML and referenced meshes are bundled, with their BSD license.
- Current vendor inventory: **355 non-cache files**, no missing/changed/extra files.
- `uv.lock`: **154 package records** including this project. The hashed pip export contains the
matching **153 external name/version pairs**; layout tests enforce this correspondence.
- New project `.venv` installed with `uv sync --locked --extra dev --python 3.13`:
Python **3.13.16**, Torch **2.14.0+cu130**, MuJoCo/MuJoCo Warp **3.16.0**, Warp **1.18.0**.
Dependency consistency and installed backend-origin checks passed; no separate mjlab/RSL-RL
distributions are installed in it.
The first large Torch download attempt timed out; a subsequent download/installation completed.
The external reference environment and source repositories were not modified. The validated
runtime does not import backend code from them.
## Automated and core-only checks
`./scripts/check`: **93 tests passed**, Ruff lint/format passed; one upstream TorchScript
warning remains. New coverage includes manual index/link/CLI synchronization, private-path and
sys.path rules, vendor drift, lock/export correspondence, installed origins outside the repo,
bad explicit interpreter/backend overrides, and negative robot-contract cases.
A separate lightweight environment installed the project with `--no-deps` plus pytest:
**76 passed, 3 skipped**. GPU/backend-dependent modules were deliberately skipped; this verifies
the core/layout CI path, not simulator functionality. The GitHub workflow is configured but has
not been executed on hosted GitHub infrastructure here.
## Actual robot gates
The setup-env check script completed all five gates—vendor, runtime, dependencies, robot and
GPU smoke—for both example configurations:
| Report | Robot checks | Probe | Peak left/right contact force | Initial penetration |
|---|---:|---|---|---:|
| `runs/setup_vendor_g1/robot.json` | 11/11 passed | 32 steps, no termination | 535.05 / 535.13 N | 0 m |
| `runs/setup_vendor_t1/robot.json` | 11/11 passed | 32 steps, no termination | 460.06 / 460.04 N | 0 m |
Forces are transient simulated contact measurements during the nominal drop/PD probe, not certified
load ratings or calibrated hardware force estimates. Sites, local frames, IMU parent/type, joint
orders, gains/limits, initial pose/velocity and timing were checked against separate declarations.
An intentionally wrong G1 declaration required a nonexistent foot site. The real CUDA gate returned
**exit 1**, `ready: false`, with **only `foot_sites` failing**; the other ten checks still ran.
Evidence: `runs/negative_robot_spec.json`, `runs/negative_robot_check.json`. This was not just a
mocked negative test. Additional unit cases corrupt IMU frames, positions, PD, order, collisions,
contact force, initial velocity, penetration and finite-state/termination behavior.
## Relocated checkout, new environment, different working directory
A copy excluding `.venv`, runs, caches and build outputs was placed under a temporary directory
whose checkout name contained a space. A **separate environment** was installed there using
`uv sync --locked --offline --extra dev --python 3.13`, using downloaded public wheel cache entries.
No original backend checkout or virtual environment was on its import path.
The subprocess environment removed `PYTHONPATH`, `MJLAB_ROOT`, `BOOSTER_ASSETS_ROOT`,
`UNITREE_ROBOTS_ROOT`, `UH_PYTHON` and `UH_EXTRA_PYTHONPATH`. From **outside that checkout**:
1. Backend origins, lock versions and CUDA passed.
2. Vendor inventory passed.
3. Bundled T1 robot gate passed.
4. T1 PPO → randomized evaluation → ONNX export completed.
5. Bundle inference verification passed.
Evidence: `runs/portability_relocation_v2.json`, `runs/portability_relocation_v2.log`,
`runs/relocated_t1_check_v2.json`, and a retained copy at `runs/relocated_t1_pipeline_v2/`.
This proves independence from the prior workspace layout on the tested host. It is not a test
on a second GPU/OS or a claim that an existing editable virtualenv can be moved without reinstalling.
## Wheel resource and installed-package checks
An initial wheel parity check caught missing `mjlab/scripts/train.py`: namespace discovery had
been disabled. Packaging now includes namespaces; a layout assertion, `tools/check_wheel.py`
and the CI build/parity step guard the regression.
The corrected wheel passed **370 exact file comparisons**, including backend Python modules,
T1/G1 model resources, license files and robot-spec JSONs. It was then installed in the relocated
environment, replacing only the editable project distribution. Imports resolved to the wheel's
`site-packages`, not either checkout. T1's actual robot gate passed again from outside the repo:
`runs/wheel_validation_v2.log`, `runs/wheel_t1_check_v2.json`.
## Training/export regression
Both final pipelines used **16 environments × 2 PPO updates × 24 steps**, followed by
**2 evaluation worlds × 150 steps**:
| Run | Failures in evaluation | ONNX max absolute error (16 actual observations) |
|---|---:|---:|
| `runs/vendored_g1_pipeline_v2` | 3 | 5.59e-8 |
| `runs/relocated_t1_pipeline_v2` | 2 | 4.66e-8 |
Both bundles passed inference verification. `runs/vendored_resume_v2` also resumed G1 for one
additional update as an installed background worker; progress reached **3 total updates**.
No validation workers were intentionally left running.
These short policies still fall. Adapter conformance, installation portability and export parity
are engineering checks—not proof of learned walking or real-robot safety. Older validation
runs retain their original source/dependency identities and require their recorded environment
for replay. Do not rewrite historical hashes after vendoring.
## Remaining release work
Original code now has an Apache-2.0 license. Review downstream attribution, set actual hosting
metadata, and validate additional hardware/driver/platform combinations. The dependency lock
cannot provide a driver, certify dynamics or substitute for a hardware safety controller.
+50
View File
@@ -0,0 +1,50 @@
# Public release checklist
A public repository description should explain the reusable framework, not assume access to a
maintainer's workstation. Documentation portability and executable portability are separate tasks.
This checklist records what is still needed before describing the project as a reproducible,
ready-to-install release.
## Packaging and dependencies
- [x] Include backend copies, required G1/T1 adapters/assets, source inventories and license notices; see `VENDOR.md`.
- [x] Publish `uv.lock`, its hashed pip export and reference CUDA/platform guidance.
- [x] Remove personal workspace defaults from launchers, T1 and asset discovery.
- [x] Reject bad explicit interpreter paths and legacy external-backend overrides.
- [x] Install the lock in a new project environment without the external backend environment.
- [x] Test a relocated checkout/new environment from another working directory, plus wheel resource parity and installed T1 conformance; see `PORTABILITY_VALIDATION.md`.
- [ ] Independently validate additional machines/GPUs and any additional platform-specific lock.
- [x] Document bundled model variants/licenses and explicit optional external overrides.
## Repository presentation
- [x] Use audience-neutral English and Chinese introductions and configurable-path command examples.
- [x] Distinguish architectural goals, implemented features, reference validation and unsupported capabilities.
- [x] Document setup requirements and current distribution limitations.
- [x] Add Apache-2.0 for original code, retain third-party licenses, and include license texts in package metadata. `NOTICE.md` remains attribution, not a replacement license.
- [ ] Set the actual repository URL and package metadata when hosting is established; do not publish placeholder links as real project URLs.
- [x] Add contribution guidance, stage manuals, synchronization tests and core/layout CI configuration.
- [ ] Configure a GPU CI runner and hosted issue templates when the repository is published.
Suggested GitHub repository description:
> An agent-ready framework for modular humanoid reinforcement learning with mjlab and MuJoCo Warp: reward iteration, training, evaluation, replay, and ONNX export.
Suggested topics: `humanoid-robotics`, `reinforcement-learning`, `mujoco`, `mjlab`, `robot-learning`,
`ai-agents`, `sim-to-real` (research direction, not a claim of hardware deployment support).
## Artifacts, privacy and claims
- [ ] Inspect the actual files staged for publication. Ignore rules do not remove files already tracked by Git.
- [ ] Exclude credentials, `.env`, virtual environments, proprietary meshes and unlicensed third-party files.
- [ ] Review run manifests, logs and archives for usernames, absolute paths, git status and source licensing before sharing them as release assets.
- [ ] Do not present development-only `runs/` paths as downloadable artifacts unless those artifacts have actually been published.
- [ ] Separate successful pipeline execution from policy convergence; include failed episodes and evaluation protocol details.
- [ ] Preserve `hardware_approved: false` and the absence of a physical robot execution interface.
## Release validation
Run lint/unit tests, selected adapter tests, GPU smoke, bounded training, fixed-protocol evaluation,
video recording and ONNX parity in the documented release environment. Record commit and dependency
identities alongside measured evidence. Existing historical validation is useful context, not a
substitute for verifying the released revision on a clean installation.
+43
View File
@@ -0,0 +1,43 @@
# Repository baseline
This directory is a standalone Git repository on `main`. The initial snapshot includes source,
vendored backends/assets, licenses, experiment recipes, manuals, tests, CI and dependency locks.
No remote is configured or pushed automatically. Configure your actual hosting URL before pushing;
do not reuse the reference project's remote.
## License
Original code and documentation use Apache-2.0, following Jumper. The unmodified license text
is in [LICENSE](../LICENSE); project attribution and third-party obligations are in
[NOTICE.md](../NOTICE.md). Existing Apache/BSD/MIT vendor licenses remain intact. Package metadata
uses the combined `Apache-2.0 AND BSD-3-Clause AND MIT` expression for the distribution, with
license texts included in the wheel.
## Locks and CI
- `uv.lock`: complete reference Linux CUDA dependency resolution, plus the lightweight CI group.
- `requirements-linux-cu130.lock`: hashed pip export for the development/training environment.
- `requirements-ci.lock`: hashed pip export for check/build tools only, without Torch/CUDA.
- `.github/workflows/check.yml`: immutable action revisions, pinned uv, locked check/build tools,
lock/export synchronization, vendor inventory, wheel parity, Ruff and core/layout tests.
The CI job does not run GPU simulation. Hosted Actions only run after publication on GitHub;
local reproduction is not a claim of a hosted green build. See [Setup](SETUP.md) for commands.
## Local verification
The full configured environment passed **94 tests** and Ruff lint/format checks. A separate
locked CI-only environment passed **77 tests, 3 skipped**; GPU/backend-dependent tests were
intentionally skipped. Wheel parity checked **370 package files** plus **7 distribution license
files** and the SPDX metadata expression. Both requirements exports were regenerated from the lock without drift.
The staged snapshot was reviewed for generated files, symlinks, oversized files, personal
workspace paths and common secret formats. `.venv*`, runs, logs, bundles, build output, caches,
checkpoints, `.env` variants and private-key files are excluded. This screening is not a guarantee
that no possible sensitive information exists; review changes before publishing.
## Provenance boundary
Existing runs remain untouched. Git initialization and lock updates are not retraining or renewed
GPU acceptance evidence. Historical runs keep their original source/lock identities; restore their
trusted recorded environment when needed rather than modifying provenance to force reuse.
+162
View File
@@ -0,0 +1,162 @@
# Environment setup
The simulation/learning sources are **vendored in this repository**:
`rl/mjlab` and `rl/rsl_rl`. G1 and T1 models required by the registered examples are also bundled,
with their licenses. No external mjlab checkout, personal directory layout or asset download is
needed for these examples. See [VENDOR.md](VENDOR.md) for exact origins and retained modifications.
## Supported reference target
The locked training target is **Linux x86-64, Python 3.11–3.13, NVIDIA CUDA 13.0 PyTorch**.
Python 3.13 is the tested interpreter. A compatible NVIDIA driver is required; the Python lock
cannot install a driver or guarantee support for every GPU. EGL is the default offscreen renderer.
Other OS/architectures and CUDA builds need a separately validated lock; there is no native-CPU
training fallback.
Before installing, the standard-library detector works without simulation packages:
```bash
python3 .agents/skills/setup-env/scripts/detect.py
```
## Recommended: locked installation in a new environment
Install `uv`, then run from the repository root:
```bash
uv sync --locked --extra dev --python 3.13
./scripts/uh doctor
./scripts/uh verify-vendor
./scripts/check
```
`uv.lock` pins transitive dependencies and artifact hashes, including the PyTorch CUDA index and
NVIDIA Warp index. `--locked` refuses to silently resolve a changed project. `uv sync` installs
this project as editable and creates `.venv`; it does not reuse another project's environment.
Large CUDA wheels may take time to download. Do not interpret a partial download as an installed
runtime.
## Standard pip editable installation
The same packaging layout supports `pip install -e .`, as in Jumper. **mjlab and rsl_rl are included
in this distribution**, not separately installed PyPI packages. Always use a fresh virtual environment:
```bash
python3.13 -m venv .venv
.venv/bin/python -m pip install -e '.[dev]'
```
This convenient command resolves the declared dependencies but does not enforce the entire
transitive lock. For the locked pip alternative, install the exported requirements first:
```bash
.venv/bin/python -m pip install --require-hashes -r requirements-linux-cu130.lock \
--extra-index-url https://download.pytorch.org/whl/cu130 \
--extra-index-url https://pypi.nvidia.com/
.venv/bin/python -m pip install --no-deps -e .
```
The exported file contains exact versions and hashes. It targets the same Linux CUDA environment,
not all platforms. Regenerate it from `uv.lock`, never edit it by hand:
```bash
uv export --locked --extra dev --no-emit-project --format requirements-txt \
--output-file requirements-linux-cu130.lock
```
Do not install `mjlab` or `rsl-rl-lib` separately. `doctor` checks package origins and detects
conflicting distributions. Repair a polluted environment by creating a new one, not by uninstalling
packages from another project's shared runtime. Installed wheels include the same model resources
and licenses; the source inventory check itself requires a repository checkout.
## Lightweight CI environment
`.github/workflows/check.yml` checks locks, licenses/layout, source inventory, wheel resource
parity, lint/format and core tests on hosted Linux/Python 3.13. Actions use immutable revisions;
the resolver is version-pinned. No GPU runtime is installed by this job, so backend-dependent
tests skip rather than claiming a simulator pass. The GPU gates below remain separate.
The `ci` dependency group is locked in `uv.lock`. Its small hashed pip export is
`requirements-ci.lock`; regenerate it with:
```bash
uv export --locked --only-group ci --no-emit-project --format requirements-txt \
--output-file requirements-ci.lock
```
CI regenerates both exports and rejects Git diffs. To reproduce its environment without replacing
your training `.venv`, use another directory:
```bash
UV_PROJECT_ENVIRONMENT=.venv-ci uv sync --locked --only-group ci --python 3.13
uv pip install --python .venv-ci/bin/python --no-deps --no-build-isolation -e .
PYTEST_DISABLE_PLUGIN_AUTOLOAD=1 .venv-ci/bin/python -m pytest -q
```
The wheel is built with `--no-isolation`, using the locked build tools, rather than downloading
an unpinned build environment. CI is triggered by pushes, pull requests and manual dispatch after
this repository is hosted on GitHub; a local Git commit does not itself execute GitHub Actions.
## Run the setup gates
```bash
.venv/bin/python .agents/skills/setup-env/scripts/check.py \
--config experiments/g1_smoke.json --output runs/setup-g1
.venv/bin/python .agents/skills/setup-env/scripts/check.py \
--config experiments/t1_smoke.json --output runs/setup-t1
```
The script runs vendor inventory, backend/CUDA/version checks, dependency consistency,
`check-robot`, and a bounded GPU smoke. It preserves a log for every gate and a JSON summary;
any failed gate gives a nonzero exit. It does not install packages, change gains, repair assets
or weaken checks. Use a new output directory each time.
`check-robot` validates declared interfaces AND measured contact response; it does not prove
locomotion or physical safety. See [NEW_ROBOT.md](NEW_ROBOT.md).
## Interpreter and optional asset configuration
`scripts/uh` chooses an explicit executable `UH_PYTHON`, otherwise the project `.venv/bin/python`,
otherwise `python3` from PATH. The project must be installed in that interpreter. A bad explicit
path fails; it never falls back to another project's environment. The launcher clears inherited
`PYTHONPATH`/user-site leakage and does not inject source directories. An installed `uh` entry
point also works from outside the repository.
`scripts/uh` rejects legacy `MJLAB_ROOT` and `UH_EXTRA_PYTHONPATH` overrides. Unset them when migrating.
The installed Python entry point also enforces vendored import origins; those old variables cannot select a backend.
Use `UH_PYTHON` only to select another **installed, compatible** environment.
Optional external asset discovery/override:
```bash
export UNITREE_ROBOTS_ROOT=/path/to/unitree_ros/robots
./scripts/uh assets
./scripts/uh inspect-asset "$UNITREE_ROBOTS_ROOT/g1_description/g1_29dof.xml"
# Optional T1 replacement, not required for the bundled example:
export BOOSTER_ASSETS_ROOT=/path/to/booster_assets
```
An explicit Booster override expects `robots/T1/T1_23dof.xml` and referenced meshes. An invalid
override fails rather than silently using the bundled asset. G1 training uses the bundled adapted
G1 model, not Unitree ROS XML. External model changes require a new conformance report and source/
asset provenance; an inventory entry is not a registered task.
## Lightweight configuration inspection
`list` and `plan` do not import the simulation stack. If intentionally inspecting source without
an install, `PYTHONPATH=src python -m universal_humanoid list` remains a metadata-only convenience;
it is **not** a supported way to select a training backend.
## Troubleshooting
- **Backend shadowing:** inspect `doctor` origins. Both backends must resolve to this checkout's
`rl/` (editable install), or the packages installed by the same built wheel.
- **Core version mismatch:** restore the documented lock in a new environment.
- **No CUDA:** check driver compatibility and GPU access; CPU discovery is not a training pass.
- **Vendor mismatch:** inspect the diff against `rl/VENDOR.json` / `rl/upstream/import.json`;
do not regenerate the inventory just to hide an unexplained change.
- **Robot gate failed:** read the failed check IDs and actual/declared values. Do not widen a
tolerance or remove a foot/sensor requirement merely to turn the report green.
- **Historical run rejected:** restore its trusted recorded source/assets/dependencies separately;
vendoring intentionally changes provenance. Do not rewrite old hashes.
+205
View File
@@ -0,0 +1,205 @@
# Training operations and experimental evidence
Training duration is not evidence of learning. Read tracking curves, exploration noise and
completed episode lengths during training; then measure the policy under explicit evaluation
conditions. These commands use the same configured runtime as the main README.
## 1. Train in the foreground or background
```bash
./scripts/uh train --config experiments/g1_flat.json --output runs/g1-baseline --background
./scripts/uh status runs/g1-baseline
./scripts/uh checkpoints runs/g1-baseline
./scripts/uh stop runs/g1-baseline
```
Omit `--background` for foreground training. Background execution is a **local Linux process**,
not a remote scheduler or reboot-persistent service. It uses the selected Python environment,
redirects stdout/stderr to `runs/g1-baseline.job/stdout.log`, and disconnects stdin/session from
the launching terminal. No TensorBoard HTTP server is started.
`status` is machine-readable JSON: process liveness, effective state, latest committed checkpoint,
completed/target iterations, session environment steps, iteration duration, learning rate,
action standard deviation, losses, last-update age and an approximate ETA. ETA includes setup
cost and is not an acceptance metric. Progress updates once per completed optimizer iteration;
a long compilation/rollout may be alive without a new progress record. A dead worker whose
manifest still says running is reported as `lost`, not completed.
`stop` sends SIGTERM to the recorded process, guarded by Linux process identity and pidfd to
avoid signaling a reused PID. Poll `status` until termination. Graceful interruption preserves
previous committed checkpoints; an incomplete optimizer update is not saved as a usable model.
SIGKILL, OOM or power loss may leave temporary files; only checkpoints published in the manifest
are usable. Failed runs and job logs are retained.
Set `checkpoint_interval` in an experiment JSON to control periodic saves (default 50 optimizer
iterations). The backend also saves after its first update and at completion. Each checkpoint
is written to a temporary file, renamed, hashed and committed to `run.json`. Published files
are immutable; new experiments and resumes always use **new output directories**.
## 2. Resume a specific checkpoint
```bash
./scripts/uh train --resume runs/g1-baseline/model_49.pt \
--iterations 200 --output runs/g1-continued --background
```
`--iterations 200` means **200 additional updates**, not a total of 200. Choose a filename from
`checkpoints`; do not assume `model_49.pt` exists. `model_N.pt` stores the zero-based last update:
`model_49.pt` represents 50 completed updates, so its continuation starts with update index 50.
The manifest records `completed_iterations` to remove this ambiguity.
Resume restores actor, critic, observation normalization, optimizer state (including learning
rate), and the backend's common-step curriculum counter. Simulator state, pending rollouts,
random-number state and logger history restart: this is **learning continuation, not bitwise
trajectory replay**. The new run records parent checkpoint and experiment hashes. Code, assets,
dependencies and experiment settings must match; only the additional iteration budget and
checkpoint interval may change. Older checkpoints lacking explicit completed-update metadata
are not accepted for continuation.
A committed intermediate checkpoint may be evaluated or resumed even if its original run is
still training or was interrupted. Export still requires a completed training run. Uncommitted
files are never recovered by guessing their modification time.
## 3. Inspect TensorBoard curves without a server
```bash
./scripts/uh curves runs/g1-baseline --output analysis/g1-curves --window 20
# Inspect another set of scalar tags:
./scripts/uh curves runs/g1-baseline --output analysis/g1-losses --match 'Loss/|Metrics/'
```
Outputs:
- `curves.json`: all selected scalar samples, available tags, missing categories, first/last
values, min/max, recent-window mean and slope per iteration.
- `curves.csv`: tag, scalar step, step-axis type, wall-clock timestamp, value.
- `curves.svg`: independently labelled panels; no GUI or plotting dependency required.
By default, select tracking terms (`Episode_Reward/track_*`, velocity-error metrics),
`Policy/mean_std`, `Train/mean_episode_length`, total reward and curriculum terms. The command
excludes RSL-RL's duplicate `/time` tags by default; if explicitly matched, these are labelled
elapsed seconds, not iterations, and their slopes are per second. It reads scalar TensorBoard
events directly in the specified run directory, including while a
training process is active. It does not recursively mix different runs. Duplicate steps use the
latest wall-time sample; non-finite points are counted and excluded, not disguised as zeros.
A missing curve is reported as missing. Episode metrics may not exist until episodes finish.
Interpret together:
- Rising total reward with flat tracking may reflect regularizer exploitation rather than command following.
- Inspect action standard deviation together with learning rate/losses; a plateau alone does not diagnose its cause.
- Check episode length, failure events, command coverage and video—not just a smooth reward curve.
- Tracking rewards are weighted training terms, not physical RMSE or calibrated success probabilities.
No curve slope automatically declares convergence or approves a policy.
## 4. Evaluate independent initial-state draws and disturbances
```bash
./scripts/uh evaluate --checkpoint runs/g1-baseline/final.pt \
--output analysis/g1-stress.json --seed 101 --profile stress \
--command 0.3 0 0 --num-envs 8 --steps 1000
```
Profiles are versioned in the recorded `fixed-twist-v2` protocol:
| Profile | Initial state on every reset | Actor observation noise | Interval disturbances |
|---|---|---|---|
| `nominal` | Fixed pose/joints/velocity | Off | None |
| `randomized` (evaluation default) | Separate uniform samples for each world | Off | None |
| `stress` | Same reset distribution as randomized | Configured task noise enabled | Per-world random velocity kicks |
Randomized reset offsets: root x/y ±0.1 m, z +0.01…0.03 m, roll/pitch ±0.03 rad, yaw ±0.2 rad;
root x/y velocity ±0.05 m/s, yaw rate ±0.1 rad/s; joint position ±0.03 rad and velocity ±0.1 rad/s.
Joint positions are clamped to the adapter's soft limits. Root offsets are relative to the robot's
default root state and world origin. These are modest test distributions, not physical validity
guarantees for future adapters.
Stress kicks add x/y velocity increments within ±0.4 m/s and yaw-rate increments within
±0.5 rad/s, at independently sampled 0.5–1.5 s intervals per world (timers restart on reset).
These are **mass-independent instantaneous velocity changes**, not force-calibrated pushes.
Training startup mass/friction/COM randomization is not enabled by these profiles; this is not a
complete sim-to-real robustness suite.
Reports include all initial states relative to world origins, unique-state counts, exact reset
ranges, actor noise definitions, per-world kick counts, failure counts and first-episode censoring.
A short rollout may experience no kick; inspect the actual counts. Repeating a seed reproduces
initial-state draws, but CUDA trajectories are not guaranteed bitwise deterministic. Distinct world samples share
a policy and are not independent training replicas. Nominal replicas remain intentionally fixed.
## 5. Run a checkpoint × seed × four-command matrix
```bash
./scripts/uh benchmark \
--checkpoints runs/g1-baseline/model_0.pt runs/g1-baseline/final.pt \
--seeds 101 202 303 --profiles randomized stress \
--steps 1000 --num-envs 8 --output analysis/g1-matrix
```
The four commands are standing `(0,0,0)`, forward `(0.3,0,0)`, lateral `(0,0.2,0)`, and turn
`(0,0,0.3)`. This example has 2 × 3 × 2 × 4 = **48 evaluation cells**, executed serially to bound
GPU memory. `summary.json` updates after every cell and records failures; completed cells are
retained, but automatic continuation of an interrupted matrix is not implemented.
Outputs include per-cell reports and aggregate `table.csv` / `table.md`:
- Mean and **sample variance across evaluation seeds**, with the number of defined values.
- XY/yaw RMSE, completed-episode duration and failure metrics.
- **Failure fraction among completed episodes** = failures / completed episodes; null if none ended.
- **Failure incidence** = failure events / simulated world-seconds, including repeated episodes.
- **First-episode failure fraction** = worlds whose first episode failed within the observation
horizon / worlds. Report censoring alongside it; unfinished episodes are not proven successes.
Variance is null with fewer than two values. Never treat repeated world resets as independent
training seeds. If a termination and timeout coincide, failure takes precedence. Means weight
seed-level reports equally; the table also supplies raw completion/failure counts. Long-lived
censored episodes can bias completed-only statistics, so inspect the entire set of metrics.
`--seeds` changes **evaluation randomness**, not training seeds. For training-seed robustness,
train separate experiment JSONs with different `seed` values, then pass their checkpoints to
this command. Results remain separated by checkpoint, with each training seed recorded.
### Select, compare, then validate on held-out seeds
```bash
./scripts/uh select-checkpoint analysis/g1-matrix/summary.json \
--metric failure_events_per_sim_second --output analysis/g1-selection.json
# Each input here must be a completed matrix containing one checkpoint:
./scripts/uh compare analysis/baseline-matrix/summary.json analysis/candidate-matrix/summary.json
```
Selection ranks the **explicit candidate list**, equally weighting command/profile mean values
for a named lower-is-better metric. No undefined cells are accepted. Ties retain input order;
there is no hidden composite quality score or hardware promotion. The selection file references
the checkpoint; it does not overwrite `final.pt`. Evaluate the winner on held-out seeds and
inspect tracking/video, since low failure incidence alone can favor standing still.
Matrix comparison reports paired candidate-minus-baseline means and sample variances by seed,
command and profile. It refuses mismatched protocols; it does not claim statistical significance.
The original two-single-report comparison remains available for quick diagnostics.
## 6. Replay with joint measurements
```bash
./scripts/uh play --checkpoint runs/g1-baseline/final.pt \
--measure analysis/g1-joints.csv --steps 1000 --seed 101 \
--profile stress --command 0.3 0 0 --video analysis/g1-replay.mp4
```
`play --measure` is a **bounded headless replay**, not an interactive viewer session. Without
`--measure`, `play --viewer native|viser` retains the interactive viewer. For multiple worlds,
use `evaluate --measure FILE --num-envs N`.
The long-form CSV has one row per controlled joint × world × control step: global and episode
time, world/episode IDs, termination/timeout flags, joint name, position (rad), velocity (rad/s),
actuator contribution to joint torque (Nm), raw/applied action and actual joint target.
Simulation uses manual resets during evaluation. CSV and video are sampled **after the control
step but before resetting ended worlds**. Torque is `qfrc_actuator` projected into joint DOFs,
not motor-space force, a desired torque, or a decimation-averaged value. Sampling is at policy
rate; substep peaks may be missed. A `.csv.json` sidecar records schema semantics, row count and
completion/failure status; `.evaluation.json` contains task metrics.
All commands remain simulation-only. These measurements do not replace hardware sensing,
transport adapters, safety supervision or sim-to-real validation.
+103
View File
@@ -0,0 +1,103 @@
# Reference-environment validation
For current vendored installation, relocated-checkout/wheel tests and automatic robot gates,
see [Portability validation](PORTABILITY_VALIDATION.md).
For the subsequent background/resume, randomized matrix evaluation, TensorBoard and CSV work,
see [Training-operations validation](OPERATIONS_VALIDATION.md). The sections below preserve the
initial pipeline validation and its original source/protocol version.
This is a historical validation record for one development configuration, not a minimum
hardware specification or a guarantee of compatibility with every installation. See
[Setup](SETUP.md) for environment requirements and [Releasing](RELEASING.md) for distribution
limitations. Paths under `runs/` below identify development artifacts; they are ignored by Git
and are not bundled with a repository checkout or offered here as downloads.
Measured on **2026-11-03**, NVIDIA GeForce **RTX 5080, 16 GB**, in a reference mjlab
Python 3.13 environment. These results validate engineering plumbing, **not learned locomotion**.
## Runtime
| Component | Observed version |
|---|---|
| mjlab | 1.6.0, customized checkout |
| mjlab git HEAD | `8bb2bc9e187797689ec82c78c35019afd2fd0376` (dirty/untracked changes retained) |
| MuJoCo / MuJoCo Warp | 3.16.0 / 3.16.0 |
| Warp | 1.18.0 |
| PyTorch | 2.14.0+cu130 |
| RSL-RL | 5.5.1 |
The exact Python/asset hashes and dirty git status are in each `train/run.json`. The reference
repositories were not modified. `sources.zip` preserves 230 Python/license entries per final run.
Do not replace the customized dependency with a public version just because its version number matches.
## Automated checks
`./scripts/check`: **35 tests passed**, Ruff lint and format checks passed. One upstream
`torch.jit.script` deprecation warning remains. Tests cover strict configs, lightweight discovery,
independent factories, reward shape/sign, unknown overrides, retained evaluation terminations,
source/experiment/checkpoint drift, partial-run rejection, bundle tampering/path traversal and
protocol mismatch.
An initial direct pytest invocation inherited ROS's Python 3.12 plugin path and failed on a missing
`lark` import. The project check script excludes inherited ROS paths and disables unrelated pytest
plugin autoload; no external Python environment was modified.
## Actual CUDA simulation and PPO
Zero-action checks ran for G1 and T1 with 4 environments and 8 steps; observations and rewards
were finite. Shapes: G1 actor/critic/action = **99 / 111 / 29**; T1 = **81 / 93 / 23**.
Final end-to-end runs each used **16 environments, 2 PPO iterations, 24 rollout steps per iteration**
(768 collected environment transitions), then **4 evaluation worlds × 500 steps**, dt 0.02 s,
5 s episode limit, fixed command `(0.3, 0, 0)`, seed 42.
| Run directory | XY tracking RMSE (m/s) | Yaw RMSE (rad/s) | Failure events | Mean completed episode (s) | ONNX max abs error |
|---|---:|---:|---:|---:|---:|
| `runs/g1_validation` | 0.709230 | 0.322807 | 24 | 1.500 | 5.59e-8 |
| `runs/t1_validation` | 0.480955 | 0.047538 | 24 | 1.560 | 5.96e-8 |
| `runs/g1_reward_validation` | 0.675789 | 0.313275 | 24 | 1.480 | 5.22e-8 |
**All three policies fall. None is a deployment candidate.** Failure events count repeated
episode failures across resets, not a probability. The current nominal evaluation deliberately
uses identical resets in parallel worlds; those worlds are not independent statistical trials.
The small candidate RMSE difference is not evidence of an improved learned policy after two updates.
Each run has `pipeline.json`, `train/final.pt`, `evaluation.json`, and `bundle/`.
`verify-bundle --inference` passed for G1 and T1. ONNX parity was checked against the PyTorch actor
at 16 actual rollout observations for every export. The reward candidate exercises local reward
function injection, training, evaluation, export and protocol-matched comparison.
## Replay evidence
`runs/t1_validation/replay.mp4`: H.264, **100 frames / 2 s / 50 FPS / 320×240**;
`video_evaluation.json` is the corresponding one-world report. Encoding and decoding succeeded,
and a decoded frame was inspected: the T1 and ground are visible, not an empty render.
This replay also records an episode failure. Desktop native and interactive Viser sessions were
not manually exercised; their CLI paths use mjlab's existing viewer classes.
## External Unitree asset checks
Native MuJoCo compilation succeeded for:
- `unitree_ros/robots/g1_description/g1_29dof.xml`: nq=36, nv=35, nu=29.
- `unitree_ros/robots/h1_description/mjcf/h1.xml`: nq=26, nv=25, nu=19.
These are **asset compilation checks only**. They are not registered RL task adapters; the
validated G1 training run uses mjlab's adapted G1 asset. T1 training uses an external Booster
T1 asset through the registered adapter.
## Failed development runs retained
`runs/g1_smoke_v1` trained but evaluation failed because the upstream command rejected
`heading_command=False` with a non-null heading range. The adapter now clears the heading range
and regression tests check evaluation configuration. `runs/g1_smoke_v2` passed an earlier short
pipeline before the final provenance/archive changes. Its old checkpoints intentionally fail
current source identity checks; use the final `*_validation` runs above.
## Remaining validation
Long-run convergence, gait quality, domain-randomized/stress evaluation, independent seeds/reset
conditions, contact realism, raw Unitree asset adaptation, rough terrain, motion imitation,
sim-to-sim dynamics parity and all physical deployment safety remain unverified/unimplemented.
No long 3,000-iteration recipe, robot transport or motor-enable operation was executed.
+79
View File
@@ -0,0 +1,79 @@
# Vendored backend sources and assets
Like Jumper's `rl/` layout, this project installs backend copies as **top-level packages** through
`pip install -e .` / `uv sync`, not through working-directory-dependent import tricks.
`rl/` deliberately has no `__init__.py`. Package discovery includes namespaces: upstream's
`mjlab/scripts` has no initializer but still contains real source that must be included in wheels.
| Path | Source | Baseline | License |
|---|---|---|---|
| `rl/mjlab/` | [mujocolab/mjlab](https://github.com/mujocolab/mjlab) plus recorded local changes | 1.6.0, commit in `rl/upstream/import.json` | Apache-2.0; embedded components retain their licenses |
| `rl/rsl_rl/` | Installed `rsl-rl-lib` release, checked against its distribution RECORD hashes | 5.5.1 | BSD-3-Clause |
| `rl/mjlab/asset_zoo/robots/booster_t1/assets/` | [BoosterRobotics/booster_assets](https://github.com/BoosterRobotics/booster_assets) | `T1_23dof.xml` and its referenced meshes; commit recorded | BSD-3-Clause |
| `rl/mjlab/asset_zoo/robots/unitree_g1/` | mjlab's adapted MuJoCo Menagerie model | Preserved with upstream mjlab assets | Unitree BSD-3-Clause |
## What was imported
The mjlab copy contains tracked package files from the validated checkout, **including its two
modified runtime files**, plus the T1 robot/task Python modules. Unrelated private/untracked
Apollo/R1 experiments, prompts, logs, environments and keyboard applications are not imported.
Other tracked upstream example assets remain intact; their presence does not register them as
supported Universal Humanoid tasks.
RSL-RL Python files were verified against the installed release's RECORD before copying. No
algorithm changes were applied. Its release METADATA and upstream mjlab pyproject are preserved
under `rl/upstream/`, without creating separately installed backend distributions.
## Explicit differences
| File | Origin / purpose |
|---|---|
| `rl/mjlab/sim/sim.py` | Retains the pre-existing local patch suppressing only LS_ITERATIONS/ITERATIONS warnings. This is **not** evidence of solver convergence; other overflow warnings remain. |
| `rl/mjlab/tasks/velocity/mdp/velocity_command.py` | Retains configurable `min_forward_velocity`; T1 uses a smaller minimum than G1. |
| `rl/mjlab/asset_zoo/robots/booster_t1/t1_constants.py` | The imported T1 adapter now defaults to bundled assets, with an explicit optional `BOOSTER_ASSETS_ROOT` override. No personal HOME fallback. |
| `rl/mjlab/tasks/velocity/config/booster_t1/` | T1 environment, PPO configuration and reward extension, previously external/untracked, now included. |
The first two inherited changes are preserved as the actual upstream-relative diff in
`rl/upstream/mjlab-local.patch`; their regression tests and the imported T1 test are retained in
`rl/upstream/tests/` as reference material. They are not automatically mixed into this project's
pytest suite. Our adapter/conformance/behavior tests exercise the supported integration.
Changed runtime files carry `[uh-vendor]` notices. The initial source inventory is
`rl/upstream/import.json`; the current distributable inventory is `rl/VENDOR.json`.
## Verify and update deliberately
```bash
python3 tools/vendor_manifest.py
./scripts/uh verify-vendor
```
Built-wheel resource parity can also be checked with
`python3 tools/check_wheel.py dist/universal_humanoid-0.2.0-py3-none-any.whl` after `uv build --wheel`.
It checks all package sources, model resources and licenses against the checkout, catching missing
namespace modules that an editable install could conceal.
Both inventory commands compare every non-cache file under `rl/` against recorded SHA-256 hashes; extra, missing and
modified files fail. No updater runs automatically. For a reviewed vendor revision:
1. Record the upstream release/commit, licenses and inherited patches.
2. Make changes in this repository only, adding a clear modification notice to changed source.
3. Update this document and review the diff against the initial import inventory.
4. Run the adapter, negative conformance, setup and GPU pipeline checks.
5. Explicitly record the accepted files with `python3 tools/vendor_manifest.py --record`.
6. If dependencies changed, regenerate `uv.lock` and its pip export, and repeat clean-install checks.
A hash inventory detects accidental drift; it does not prove safety or authenticate arbitrary code.
A deliberate hash refresh without review defeats its purpose. Source snapshots/provenance now
include RSL-RL source identity, robot declarations and the dependency lock, not just mjlab.
## Licensing and distribution
License texts travel inside both package directories and are included as wheel resources. The
mjlab assets and Isaac Lab-derived helpers carry their corresponding license texts in
`rl/mjlab/licenses/`; RSL-RL's dependencies and pfrl attribution are under `rl/rsl_rl/licenses/`.
T1 assets carry their own `assets/LICENSE`. See [NOTICE.md](../NOTICE.md) for attribution.
The upstream licenses do not automatically choose a license for this project's original code or
settle authorship of downstream additions. Select that license and review downstream attribution
before announcing an open-source release. No credentials or proprietary robot transport code is
included, and no physical deployment support is implied.
+15
View File
@@ -0,0 +1,15 @@
{
"task": "g1.flat",
"seed": 42,
"num_envs": 1024,
"iterations": 3000,
"steps_per_env": 24,
"episode_seconds": 20.0,
"device": "cuda:0",
"action_clip": 5.0,
"reward_weights": {
"track_linear_velocity": 2.0,
"track_angular_velocity": 2.0,
"action_rate_l2": -0.1
}
}
+19
View File
@@ -0,0 +1,19 @@
{
"task": "g1.flat",
"seed": 42,
"num_envs": 1024,
"iterations": 3000,
"steps_per_env": 24,
"episode_seconds": 20.0,
"device": "cuda:0",
"action_clip": 5.0,
"reward_weights": {"track_linear_velocity": 0.0},
"extra_rewards": {
"track_planar": {
"function": "planar_velocity_exp",
"weight": 2.0,
"params": {"std": 0.5}
},
"joint_speed": {"function": "joint_speed_l2", "weight": -0.0001}
}
}
+16
View File
@@ -0,0 +1,16 @@
{
"task": "g1.flat",
"seed": 42,
"num_envs": 16,
"iterations": 2,
"steps_per_env": 24,
"episode_seconds": 5.0,
"device": "cuda:0",
"reward_weights": {"track_linear_velocity": 0.0},
"extra_rewards": {
"track_planar": {
"function": "planar_velocity_exp", "weight": 2.0, "params": {"std": 0.5}
},
"joint_speed": {"function": "joint_speed_l2", "weight": -0.0001}
}
}
+9
View File
@@ -0,0 +1,9 @@
{
"task": "g1.flat",
"seed": 42,
"num_envs": 16,
"iterations": 2,
"steps_per_env": 24,
"episode_seconds": 5.0,
"device": "cuda:0"
}
+15
View File
@@ -0,0 +1,15 @@
{
"task": "t1.flat",
"seed": 42,
"num_envs": 1024,
"iterations": 3000,
"steps_per_env": 24,
"episode_seconds": 20.0,
"device": "cuda:0",
"action_clip": 5.0,
"reward_weights": {
"track_linear_velocity": 3.0,
"track_angular_velocity": 1.5,
"action_rate_l2": -0.03
}
}
+9
View File
@@ -0,0 +1,9 @@
{
"task": "t1.flat",
"seed": 42,
"num_envs": 16,
"iterations": 2,
"steps_per_env": 24,
"episode_seconds": 5.0,
"device": "cuda:0"
}
+30
View File
@@ -0,0 +1,30 @@
Copyright (c) 2022-2025, The Isaac Lab Project Developers (https://github.com/isaac-sim/IsaacLab/blob/main/CONTRIBUTORS.md).
All rights reserved.
SPDX-License-Identifier: BSD-3-Clause
Redistribution and use in source and binary forms, with or without modification,
are permitted provided that the following conditions are met:
1. Redistributions of source code must retain the above copyright notice,
this list of conditions and the following disclaimer.
2. Redistributions in binary form must reproduce the above copyright notice,
this list of conditions and the following disclaimer in the documentation
and/or other materials provided with the distribution.
3. Neither the name of the copyright holder nor the names of its contributors
may be used to endorse or promote products derived from this software without
specific prior written permission.
THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND
ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE FOR
ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
(INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND
ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
(INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+21
View File
@@ -0,0 +1,21 @@
MIT License
Copyright (c) 2025 i2rt robotics
Permission is hereby granted, free of charge, to any person obtaining a copy
of this software and associated documentation files (the "Software"), to deal
in the Software without restriction, including without limitation the rights
to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
copies of the Software, and to permit persons to whom the Software is
furnished to do so, subject to the following conditions:
The above copyright notice and this permission notice shall be included in all
copies or substantial portions of the Software.
THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
SOFTWARE.
+27
View File
@@ -0,0 +1,27 @@
Copyright (c) 2016-2023 HangZhou YuShu TECHNOLOGY CO.,LTD. ("Unitree Robotics")
All rights reserved.
Redistribution and use in source and binary forms, with or without
modification, are permitted provided that the following conditions are met:
* Redistributions of source code must retain the above copyright notice, this
list of conditions and the following disclaimer.
* Redistributions in binary form must reproduce the above copyright notice,
this list of conditions and the following disclaimer in the documentation
and/or other materials provided with the distribution.
* Neither the name of the copyright holder nor the names of its
contributors may be used to endorse or promote products derived from
this software without specific prior written permission.
THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE
FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+29
View File
@@ -0,0 +1,29 @@
BSD 3-Clause License
Copyright (c) 2016-2022 HangZhou YuShu TECHNOLOGY CO.,LTD. ("Unitree Robotics")
All rights reserved.
Redistribution and use in source and binary forms, with or without
modification, are permitted provided that the following conditions are met:
* Redistributions of source code must retain the above copyright notice, this
list of conditions and the following disclaimer.
* Redistributions in binary form must reproduce the above copyright notice,
this list of conditions and the following disclaimer in the documentation
and/or other materials provided with the distribution.
* Neither the name of the copyright holder nor the names of its
contributors may be used to endorse or promote products derived from
this software without specific prior written permission.
THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE
FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+21
View File
@@ -0,0 +1,21 @@
MIT License
Copyright (c) 2020 Preferred Networks, Inc.
Permission is hereby granted, free of charge, to any person obtaining a copy
of this software and associated documentation files (the "Software"), to deal
in the Software without restriction, including without limitation the rights
to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
copies of the Software, and to permit persons to whom the Software is
furnished to do so, subject to the following conditions:
The above copyright notice and this permission notice shall be included in all
copies or substantial portions of the Software.
THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
SOFTWARE.
+74
View File
@@ -0,0 +1,74 @@
[build-system]
requires = ["setuptools>=77,<83"]
build-backend = "setuptools.build_meta"
[project]
name = "universal-humanoid"
version = "0.2.0"
description = "An agent-ready framework for modular humanoid reinforcement learning with mjlab and MuJoCo Warp"
readme = "README.md"
# Original code: Apache-2.0. The wheel also redistributes BSD/MIT components.
license = "Apache-2.0 AND BSD-3-Clause AND MIT"
license-files = ["LICENSE", "NOTICE.md", "licenses/*.txt"]
requires-python = ">=3.11,<3.14"
# mjlab and rsl_rl are vendored packages, NOT separately installed distributions.
dependencies = [
"torch==2.14.0", "mujoco==3.16.0", "mujoco-warp==3.16.0", "warp-lang==1.18.0",
"numpy", "tensordict>=0.7", "torchrunx>=0.3.4", "tyro>=1.0.1", "trimesh>=4.8.3",
"scipy>=1.15", "viser>=1.0.27", "mjviser>=0.0.14", "mediapy>=1.2.7",
"imageio", "imageio-ffmpeg", "prettytable", "tqdm", "PyYAML", "GitPython>=3.1.59",
"tensorboard>=2.20", "onnx", "onnxscript>=0.5.4", "onnxruntime>=1.19",
"wandb>=0.22.3",
]
[project.optional-dependencies]
dev = ["pytest>=8", "ruff>=0.9", "build", "setuptools>=77,<83"]
# Compatibility extras; export/analysis are now included in the default installation.
export = []
analysis = []
[project.scripts]
uh = "universal_humanoid.cli:main"
# Same tools as dev, without the project/CUDA dependencies on hosted CI runners.
[dependency-groups]
ci = ["pytest>=8", "ruff>=0.9", "build", "setuptools>=77,<83"]
[tool.setuptools.packages.find]
where = ["src", "rl"]
include = ["universal_humanoid*", "mjlab*", "rsl_rl*"]
namespaces = true
[tool.setuptools.package-data]
universal_humanoid = ["robot_specs/*.json"]
mjlab = ["**/*.xml", "**/*.stl", "**/*.STL", "**/*.obj", "**/*.png", "**/*.jpg", "**/*.jpeg", "**/*.json", "**/*.txt", "**/*.md", "**/LICENSE*", "LICENSE", "py.typed"]
rsl_rl = ["LICENSE", "**/*.txt", "**/LICENSE*", "*.md"]
[tool.pytest.ini_options]
pythonpath = ["src"]
testpaths = ["tests"]
markers = ["integration: requires the installed vendored mjlab environment and assets"]
[tool.ruff]
line-length = 100
extend-exclude = ["rl/mjlab", "rl/rsl_rl", "rl/upstream"]
[tool.ruff.lint]
select = ["E", "F", "I"]
[tool.uv]
environments = ["sys_platform == 'linux' and platform_machine == 'x86_64'"]
[[tool.uv.index]]
name = "pytorch-cu130"
url = "https://download.pytorch.org/whl/cu130"
explicit = true
[[tool.uv.index]]
name = "nvidia"
url = "https://pypi.nvidia.com/"
explicit = true
[tool.uv.sources]
torch = { index = "pytorch-cu130" }
warp-lang = { index = "nvidia" }
+37
View File
@@ -0,0 +1,37 @@
# This file was autogenerated by uv via the following command:
# uv export --locked --only-group ci --no-emit-project --format requirements-txt --output-file requirements-ci.lock
build==1.6.1 ; platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:51cc11666391ab6f092070437ac747002ff46f3e4113a3622177ee6b488bfc53 \
--hash=sha256:ecd351a4be9d35a9eaaba244a7687143c9c7d4aea6ac964e7e7ddab20cbcf4e7
iniconfig==2.3.1 ; platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:67f4b9c50da0dedf52af349e7749a80a9057a5031199791b906c3bb3ae878960 \
--hash=sha256:9121e2c1fdb355232495be3194c8dfe87ccc2d5dee45947b78e68f499790d7a7
# via pytest
packaging==26.3 ; platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:94edc256424af38762eb31306eed28beb9f0efc50a8837492c9d6fd6004aed79 \
--hash=sha256:d7193f7c8e4e93f444fde0262bf90af30e16fa0ad0ad44cb553c87339b23cd1c
# via
# build
# pytest
pluggy==1.7.0 ; platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:7dd7b0d8832ba3cb632c306926ded123429211b83641b35dc5c41ad2d34f9bec \
--hash=sha256:d1eaa46ebb595891b860ab086b4d09c8588af65ebd4361b8e8f4bb8920b90ba8
# via pytest
pygments==2.21.0 ; platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:2363c69b61c4a97c838da3b130dcd6468f4848992b21a82f2a63ec34377137d9 \
--hash=sha256:610ca751c9bc2492b38eb9a38a7fbc93edbbb2d7182edaf34e66ae493dee5c8c
# via pytest
pyproject-hooks==1.3.3 ; platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:5fc53fdac9f7bd63fbcdc868fb5f90b4784d78a53a3d3388cd738b807441a20b \
--hash=sha256:defda19b854fa0d3bd4f76ea4ddcba8abd7dcfcdd585a6690ade050744fc5f43
# via build
pytest==9.2.0 ; platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:07acdcf88ec174f907dc01af70bf52d49b014bbab03faeeade31c96b18c2e588 \
--hash=sha256:97c4f3b1f3c7414aa36465e09b8220f1d03627993bd51b1282e193753c80557b
ruff==0.17.3 ; platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:692fdb46d0fd9c71e380947dc5fc72f2c3f881073bc1f739e4579376231affbe \
--hash=sha256:7fa4b0f931cdbcb8a52044f370a3eb429fb42eb88807b0e800fcaa98dd42b24e \
--hash=sha256:a57049c8addb3662073859bb55a7ad35aa54f8a7c9ed268911d1333c5aadaaa9
setuptools==82.0.1 ; platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:7d872682c5d01cfde07da7bccc7b65469d3dca203318515ada1de5eda35efbf9 \
--hash=sha256:a59e362652f08dcd477c78bb6e7bd9d80a7995bc73ce773050228a348ce2e5bb
+952
View File
@@ -0,0 +1,952 @@
# This file was autogenerated by uv via the following command:
# uv export --locked --extra dev --no-emit-project --format requirements-txt --output-file requirements-linux-cu130.lock
absl-py==2.5.1 ; platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:286e71c82c1a38e75bbcf185f9b37d0305ad7786535107cb49bf4df9ff2e1f95 \
--hash=sha256:721200f2f0e9960f2ca9dc3a2a706b201f5f75d812c158f059cbbe29eeafbdb8
# via
# mujoco
# mujoco-warp
# tensorboard
annotated-types==0.8.0 ; platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:13b2beaad985e05e2d6407ee4c4f35590b11f8d693a258a561055cac8f64cab7 \
--hash=sha256:f072f4d804ea359e4eaf198b1af7a8b0943881a87f31bb764f8bf219bb9419e0
# via pydantic
anyio==4.16.0 ; platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:59bb03251907983d1603d27c3087ddb1e93299ea2163adcdba37899a4fb12ffb \
--hash=sha256:c66ae634e0b8228396b801741beb3294326d4b6102f7bba4041c0dfb58c29026
# via
# cwevals
# httpx
asttokens==3.0.2 ; platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:3ecdbd8f2cc195f53ccada3a613538bb5f9ef6f6869129f13e03c30a677b8fe2 \
--hash=sha256:9da13157f5b28becde0bd374fc677dcd3c290614264eff096f167c469cd9f933
# via stack-data
bcrypt==5.0.0 ; platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:046ad6db88edb3c5ece4369af997938fb1c19d6a699b9c1b27b0db432faae4c4 \
--hash=sha256:0c8e093ea2532601a6f686edbc2c6b2ec24131ff5c52f7610dd64fa4553b5464 \
--hash=sha256:137c5156524328a24b9fac1cb5db0ba618bc97d11970b39184c1d87dc4bf1746 \
--hash=sha256:212139484ab3207b1f0c00633d3be92fef3c5f0af17cad155679d03ff2ee1e41 \
--hash=sha256:3cf67a804fc66fc217e6914a5635000259fbbbb12e78a99488e4d5ba445a71eb \
--hash=sha256:560ddb6ec730386e7b3b26b8b4c88197aaed924430e7b74666a586ac997249ef \
--hash=sha256:5feebf85a9cefda32966d8171f5db7e3ba964b77fdfe31919622256f80f9cf42 \
--hash=sha256:611f0a17aa4a25a69362dcc299fda5c8a3d4f160e2abb3831041feb77393a14a \
--hash=sha256:61afc381250c3182d9078551e3ac3a41da14154fbff647ddf52a769f588c4172 \
--hash=sha256:6b8f520b61e8781efee73cba14e3e8c9556ccfb375623f4f97429544734545b4 \
--hash=sha256:7aeef54b60ceddb6f30ee3db090351ecf0d40ec6e2abf41430997407a46d2254 \
--hash=sha256:89042e61b5e808b67daf24a434d89bab164d4de1746b37a8d173b6b14f3db9ff \
--hash=sha256:baade0a5657654c2984468efb7d6c110db87ea63ef5a4b54732e7e337253e44f \
--hash=sha256:d8d65b564ec849643d9f7ea05c6d9f0cd7ca23bdd4ac0c2dbef1104ab504543d \
--hash=sha256:ed2e1365e31fc73f1825fa830f1c8f8917ca1b3ca6185773b349c20fd606cec2 \
--hash=sha256:f632fd56fc4e61564f78b46a2269153122db34988e78b6be8b32d28507b7eaeb \
--hash=sha256:f748f7c2d6fd375cc93d3fba7ef4a9e3a092421b8dbf34d8d4dc06be9492dfdd \
--hash=sha256:f8429e1c410b4073944f03bd778a9e066e7fad723564a52ff91841d278dfc822
# via paramiko
build==1.6.1 ; platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:51cc11666391ab6f092070437ac747002ff46f3e4113a3622177ee6b488bfc53 \
--hash=sha256:ecd351a4be9d35a9eaaba244a7687143c9c7d4aea6ac964e7e7ddab20cbcf4e7
# via universal-humanoid
certifi==2026.10.23 ; platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:a9895e6e2758b371e0545e321a3a8fd1fcfd9829ada8824a936dfb16aaaf36cd \
--hash=sha256:e9b54de76e530bb6cf0ccd036787103acb58fd42ba930ea8bf948ddbaa2f61d1
# via
# httpcore
# httpx
# requests
cffi==2.1.1 ; platform_machine == 'x86_64' and platform_python_implementation != 'PyPy' and sys_platform == 'linux' \
--hash=sha256:210019b6c7cf07f081b4c54635c8cf744377001350e29cc0f81c4377b4797735 \
--hash=sha256:34e261f78cb6ceaaa36f42f2613f4380d94d9c759a9c73c769ee6e0247364632 \
--hash=sha256:a931079504ecc49efed7744c476a5c343a92fabf66dec2db95edb1b2fdc770e2 \
--hash=sha256:c1453022f490d2459a11819d83ad1d586e9ff65a12ac3e705ffebd46d3685dcf \
--hash=sha256:dd31f52ea1086513bb9df30f8fcee9b8918323ae067a3d5b78bc826a000712be \
--hash=sha256:e0bcb7e0f677f543555d2adff3bf19c05f66cdb4796e5ff602442ab2fe3c4ef7 \
--hash=sha256:f5cfbc5fe74540d335175b656c725d74d90e3730c626d92575eea35029d9afaa
# via
# cryptography
# pynacl
charset-normalizer==3.5.2 ; platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:114e4d0c92d618409ed82a99e22b5c5e768fe995f2973f78265f4524f49d4640 \
--hash=sha256:1c50fe28bbc2ced33386f298650d91218076c05420e6cbd790b913adc41659e7 \
--hash=sha256:211d5a3eb6af8f513b8d4ca19a8c1b7accab1b5f0d3175f9826b03c1a920dc1f \
--hash=sha256:23851fb4e1b85ed3f6c2a27b777cdfe2e19fb5b38429a8faf38c7542b7665869 \
--hash=sha256:39de2a259fc954455c57274dc94c79d5842774e1247a016aff30bc0efed0f4ef \
--hash=sha256:3d31298449090ab8d47b7b1b2a555ff73cac7ed438a08b7ac160980c7ebed649 \
--hash=sha256:7218e8f32b0956cfcd048fd42d9d5779809745ca1d86113ca56f66e7ae1549c4 \
--hash=sha256:9cae88599c7219005d879f98e5ed53341e9a122af585e1091200358a3003d2a0 \
--hash=sha256:a090bb2c68df85450502e3e20d665e3a5af9c65a84d6508ed477badd49166fd3 \
--hash=sha256:b6b751274acb69d77b3323d6b7dbaa3c7fdfc1eb829b7eb61d262f32e1af9685 \
--hash=sha256:e243bd13217235fc7290c621941c3f5cc8b66e4872495be821d7436ba2fb838d
# via requests
click==8.5.0 ; platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:255bc9599cf7748b4b1a446ccc735421bd08a2ae529a8b88597d3de5664ee360 \
--hash=sha256:ba0d2089de75ea0310e2dde03160e6ca10009947fb95a182f9b54021bb272e34
# via wandb
cloudpickle==3.1.2 ; platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:7fda9eb655c9c230dab534f1983763de5835249750e85fbcef43aaa30a9a2414 \
--hash=sha256:9acb47f6afd73f60dc1df93bb801b472f05ff42fa6c84167d25cb206be1fbf4a
# via
# tensordict
# torchrunx
contourpy==1.3.3 ; python_full_version < '3.12' and platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:083e12155b210502d0bca491432bb04d56dc3432f95a979b429f2848c3dbe880 \
--hash=sha256:2b7e9480ffe2b0cd2e787e4df64270e3a0440d9db8dc823312e2c940c167df7e \
--hash=sha256:322ab1c99b008dad206d406bb61d014cf0174df491ae9d9d0fac6a6fda4f977f \
--hash=sha256:459c1f020cd59fcfe6650180678a9993932d80d44ccde1fa1868977438f0b411 \
--hash=sha256:4d00e655fcef08aba35ec9610536bfe90267d7ab5ba944f7032549c55a146da1 \
--hash=sha256:4debd64f124ca62069f313a9cb86656ff087786016d76927ae2cf37846b006c9 \
--hash=sha256:50ed930df7289ff2a8d7afeb9603f8289e5704755c7e5c3bbd929c90c817164b \
--hash=sha256:51e79c1f7470158e838808d4a996fa9bac72c498e93d8ebe5119bc1e6becb0db \
--hash=sha256:5ed3657edf08512fc3fe81b510e35c2012fbd3081d2e26160f27ca28affec989 \
--hash=sha256:ca0fdcd73925568ca027e0b17ab07aad764be4706d0a925b89227e447d9737b7
# via matplotlib
contourpy==1.4.0 ; python_full_version >= '3.12' and platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:20156f5a1ac4f8ce02656e39a61e82164a3d359796dc8026f75b062783d500e1 \
--hash=sha256:3863ef2e2b13fe93f8c0ebb08ee400cb07153b8b0e91c5acb26e4537f283634f \
--hash=sha256:6ced1670fafee703b8277ab1645072a226e74f167166eae146fb743916119b67 \
--hash=sha256:875f42444c9cf48d56f724f2637e60d0f73b3b12c9041e1484580a233edf9591 \
--hash=sha256:abd0f7e51ecc52bf8c95713bd80b1c17c516a71fa391c54475308a4377903c2b \
--hash=sha256:f1219a8898523cba821085f2da8a1b962cc696325763f09d7f93538ba43b2d70
# via matplotlib
cryptography==50.0.2 ; platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:1981f1db4630889b9ef7803fadef12b056f428cb6b85c27ba57b774793b6093c \
--hash=sha256:1ba34f04897fcdaa73f74145c25f3ec146fbd56593853e88adc2e811303c5f42 \
--hash=sha256:4061c0079120205fb760c58acab6443e217307dcf05e3702cf970e0689972856 \
--hash=sha256:58a0c478eeca76fe5e07993c5a0703def34a6dc6a0cda4f5564639b33112ffe7 \
--hash=sha256:630ebfea3bf689d075f82316324ff7433dc447fe6bc1bfc76524b74b4a9567d2 \
--hash=sha256:7b46165bb56eb4704e2eaaf86f3c940d19154535d9b0ca7d6d590b04060e00d5 \
--hash=sha256:85d0d9a31b9098e98534226d5686b47264b95e62ce459dc2e62fdfc809f9fe93 \
--hash=sha256:9dab55f57c74c3cad24c323bacbbd04be4705ba6eb0d92e920b1fc4837ed5079 \
--hash=sha256:c71be1cbfa5cd9a41ee452acf1eccd82b2c05950358b106ec8ceb83411d1a020 \
--hash=sha256:ce47f66801c20ec6c6632453bb5960fe38939e9306970b48b3a5a26de7745d94 \
--hash=sha256:e105ab60406787da31fccc883fc0f733af1efd78f0136a4599692c4083a73d0c \
--hash=sha256:ee247f5c245c9a2fe7c8e2214e295918838e44e00a45a6718451e4004219e767 \
--hash=sha256:eef4c2f3423810b3070ab391f85436d2f8bbfcb286ac15cbc73190b3563b1f1a \
--hash=sha256:f21e8a22c8605750c7af886bab299a363721264061b4ac0a30efb73cfd58efc5 \
--hash=sha256:f2f9bd7f90c64fe89253f0a2c05e3c4856072660429ce8831b4235bf29403a67
# via paramiko
cuda-bindings==13.4.4 ; platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:433efa31e33c2868621b5f6036bd3bf5de9c45c5546cb1361a5f6ff3415b8bb5 \
--hash=sha256:5d6adc44f098472d8ac2d6211923eea6081830521852dc77f8c9237bf9320a59 \
--hash=sha256:e1cef1eaf5a08cc3bf3675ca9d3f97f3fc7085998556c848b4053e74ba4ba392
# via torch
cuda-pathfinder==1.8.5 ; platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:16d1026fdd78ac0222b966441de53dec3f0dbe6b1e2c43d19d92eb29bfd81e67
# via cuda-bindings
cuda-toolkit==13.0.3.0 ; platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:d693caaa261214ddd7dbb60d68e71cbed884e68c2be7509778f3051da0b91c3f
# via torch
cwevals==0.0.1b1 ; platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:2ae997369ff44941925e4db1bd07eb4e13c05de412e615636fd12881d63f8a13 \
--hash=sha256:6936baf0ccdc5d9092abd5c2ab628cedcdae9bb72edfdf730f7aaac19118855a
# via wandb
cycler==0.12.1 ; platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:85cef7cff222d8644161529808465972e51340599459b8ac3ccbac5a854e0d30 \
--hash=sha256:88bb128f02ba341da8ef447245a9e138fae777f6a23943da4540077d3601eb1c
# via matplotlib
decorator==5.3.1 ; platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:4cbcdd55a6efadb9dbea26b858f4fb3264567b52d69ca0d25b721b553f60ea82 \
--hash=sha256:f47fe6fdbd2edd623ecfe36875d37aba411624e2670dd395dddae1358689bb3c
# via fabric
deprecated==1.3.1 ; python_full_version < '3.12' and platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:597bfef186b6f60181535a29fbe44865ce137a5079f295b479886c82729d5f3f \
--hash=sha256:b1b50e0ff0c1fddaa5708a2c6b0a6588bb09b892825ab2b214ac9ea9d92a5223
# via fabric
deprecated==3.0.1 ; python_full_version >= '3.12' and platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:2f5a9fecc9961e26462846cf1e57f73a612b1d0b79c27dceeed8de647beadcee \
--hash=sha256:eda314c115e8bc0fe52b90a6869e19240ac3cedb7449adfdba5981cf9ef693e1
# via fabric
distro==1.9.0 ; platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:2fa77c6fd8940f116ee1d6b94a2f90b13b5ea8d019b98bc8bafdcabcdd9bdbed \
--hash=sha256:7bffd925d65168f85027d8da9af6bddab658135b840670a223589bc0c8ef02b2
# via cwevals
docstring-parser==0.18.0 ; platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:292510982205c12b1248696f44959db3cdd1740237a968ea1e2e7a900eeb2015 \
--hash=sha256:b3fcbed555c47d8479be0796ef7e19c2670d428d72e96da63f3a40122860374b
# via tyro
etils==1.15.0 ; platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:7a7e242b299813f454d3c2910b8428fd5758ef3c6df58af4808fd0a511f0a032 \
--hash=sha256:b9e7f3d7ab980c564245cd54e2b928e6abcb0521d6042ad3fdb13eea725309f0
# via
# mujoco
# mujoco-warp
executing==2.3.0 ; platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:15919cb5d667e5cb4e099511971d00d659573fff2dd5c4e6cd8b71636c7858d2 \
--hash=sha256:736e859c9f8701f11fcf516856f26f562e04776387824b43a35a1dfe21c84122
# via stack-data
fabric==3.2.3 ; platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:ce61917f4f398018337ce279b357650a3a74baecf3fdd53a5839013944af965e \
--hash=sha256:dcbd2c47ad87688facaef5cc11aab6d1ec9ed05645fed97a5de7204d5d17cc44
# via torchrunx
filelock==4.2.7 ; platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:31f6f7e3a8b6f0bd89114d30c5d5218ea1cfb63e5c4aa695b7559314a209bc7f \
--hash=sha256:d08972bbca2b8767ee552aff877b091dd111ffb1d22de7de885766ea7d719d5e
# via torch
flatbuffers==25.12.19 ; platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:7634f50c427838bb021c2d66a3d1168e9d199b0607e6329399f04846d42e20b4
# via onnxruntime
fonttools==4.68.0 ; platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:08851f7d2b983412f490b73a7c56f17cbda1d4d85d2ad3f857564bf15b5d9475 \
--hash=sha256:3edce69ecdefa73ab1041901ca22eefed6ebb93a77bc77e889d940479f25bf7d \
--hash=sha256:4b9c89b5876f1a74b1341632dfaf4c4c2be565d76c060445df401ebe7a4ceefa \
--hash=sha256:73294e233070f71e2b3653b6bd8564af676d57f26d075761fb06d1ab2e4a7a5d \
--hash=sha256:7ab8097e7e191784974c5608eb95fb31e1243fe3054cb66778ac82558e9b36f8 \
--hash=sha256:b5a015b322275d3d745b06d5e69cf74ef4b1ce2dae02afbe4826cb186d8527a3 \
--hash=sha256:d76f8fdcb6936f5e1045a0c9dd721a4bf11112e60d1e243bfa1d39f06a8b3104 \
--hash=sha256:e3ecdb0b248cd0bf43100a8b3cb40f7b45d52ea4be8616edea5dea8fefe90229
# via matplotlib
fsspec==2026.9.0 ; platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:0f08147951c8cb31d844c3547d631053b127863b60be04cf06e121333ee0e2fe \
--hash=sha256:8dd6e646e99ea382bd85f97a45e6b526a442d79423a7dc673f1e2756d05fcb5f
# via
# etils
# torch
gitdb==4.0.12 ; platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:5ef71f855d191a3326fcfbc0d5da835f26b13fbcba60c32c21091c349ffdb571 \
--hash=sha256:67073e15955400952c6565cc3e707c554a4eea2e428946f7a4c162fab9bd9bcf
# via gitpython
gitpython==3.2.1 ; platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:0f1229ec5c057bdabdeeda8bda4f17bda679ab33dbccc00f24e6a480553c8613 \
--hash=sha256:d66b0832aa8755c1a84892937282df0befb16c97e5a64aeebd68be9c5e323f53
# via universal-humanoid
glfw==2.11.0 ; platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:4bd3a462ed02ae25455dceaeb76d154faf08b48962c0cb2cf685419a34a71305 \
--hash=sha256:76484fae4a4ef0f778a05e02422ecbe6960ae23039087fe2e0ad9742ab8c7bbe \
--hash=sha256:a08e1135d02c9408d7622db3d391e3e33e4187754ec45633e9a482c775f4846c
# via mujoco
googleapis-common-protos==1.76.0 ; platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:59e13344a9f3651aa201a03e2bee815a3bf43a4baf1f9ff8a9b62a840ce25fac \
--hash=sha256:71db0e44a1d4c2dd024ebad44ca67129cc1e2892e4861df466764749d5ef8642
# via opentelemetry-exporter-otlp-proto-http
grpcio==1.85.1 ; platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:000999ea9bba80217b1625782d638e7a688790f6154e845e18826f7da859bfc7 \
--hash=sha256:5f530287d5de5f7f66089ead81de44f6787162944dab3922efa65dfdff108a1d \
--hash=sha256:83aced2a1c05648cee0de34e531351118b14c58238f0a61b43559ac8851647a8 \
--hash=sha256:83b09ca3f9602caf4f1b4295b4bb002b10477789ff9797feeedc811b9a6be0b7 \
--hash=sha256:9998f03ac0dfed226eead6e93cee1b547d7a32e47e1915dee94eff0e113b167e \
--hash=sha256:a1a72d19c0ab84b4b73b6c47eeaade76d84b1973a51593ff1160a73df33d03d9 \
--hash=sha256:de520b1555ae445cdc4f8f1cc7ccba12a19992bfafa7c292ebe7fe1fe5da6c53
# via tensorboard
h11==0.16.0 ; platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:4e35b956cf45792e4caa5885e69fba00bdbc6ffafbfa020300e549b208ee5ff1 \
--hash=sha256:63cf8bbe7522de3bf65932fda1d9c2772064ffb3dae62d55932da54b31cb6c86
# via httpcore
httpcore==1.0.9 ; platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:2d400746a40668fc9dec9810239072b40b4484b640a8c38fd654a024c7a1bf55 \
--hash=sha256:6e34463af53fd2ab5d807f399a9b45ea31c3dfa2276f15a2c3f00afff6e176e8
# via httpx
httpx==0.28.1 ; platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:75e98c5f16b0f35b567856f597f06ff2270a374470a5c2392242528e3e3e42fc \
--hash=sha256:d909fcccc110f8c7faf814ca82a9a4d816bc5a6dbfea25d6591d6985b8ba59ad
# via cwevals
idna==3.20 ; platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:a7db850025b95ded1eae8a46181a1a6c56c92c96f0e2b005d9ff8dc0210cab44 \
--hash=sha256:ab7ae7122974553370f0bdb919e1a960b2cd1bc1ef0276416d896db81c14582c
# via
# anyio
# httpx
# requests
imageio==2.38.1 ; platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:36d23eb7423d2fb63f637098758edb3d2df3687f7125e0a5ce8596572022dcf9 \
--hash=sha256:6769f1f01c4dd46448307863a787c9a22fa4dbe11c0c88f525c6e410fdfd7983
# via
# universal-humanoid
# viser
imageio-ffmpeg==0.8.0 ; platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:0385d9bab7dc7d7141af7d0f160059e05f0d874116faf9ce91f6fcfd66e938f2 \
--hash=sha256:543c8c5d316fc48575dfb5d44adba84a473d9c5a0e482b7fac5988f24a9cb17b
# via universal-humanoid
iniconfig==2.3.1 ; platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:67f4b9c50da0dedf52af349e7749a80a9057a5031199791b906c3bb3ae878960 \
--hash=sha256:9121e2c1fdb355232495be3194c8dfe87ccc2d5dee45947b78e68f499790d7a7
# via pytest
invoke==2.2.1 ; platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:2413bc441b376e5cd3f55bb5d364f973ad8bdd7bf87e53c79de3c11bf3feecc8 \
--hash=sha256:515bf49b4a48932b79b024590348da22f39c4942dff991ad1fb8b8baea1be707
# via
# fabric
# paramiko
ipython==9.18.1 ; platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:6319c0f5372c196397899479e6db1783dd7937abae06815d1317e8d5d51aaaaf \
--hash=sha256:997ce9140d374531a61069c70c0f9ef299d6dd3559ea0f94b8af589726e3c1d5
# via mediapy
ipython-pygments-lexers==1.1.1 ; platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:09c0138009e56b6854f9535736f4171d855c8c08a563a0dcd8022f78355c7e81 \
--hash=sha256:a9462224a505ade19a605f71f8fa63c2048833ce50abc86768a0d81d876dc81c
# via ipython
jedi==0.20.1 ; platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:0fb16d86c4a4c73c37ba518c77419975e30fcc620658a8d14fbb5720cdd34142 \
--hash=sha256:2f71208c3f9c1bca057c0e90d3f272aba44ace88fc4067d7587e9e069331b7e5
# via ipython
jinja2==3.1.6 ; platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:0137fb05990d35f1275a587e9aee6d56da821fc83491a0fb838183be43f66d6d \
--hash=sha256:85ece4451f492d0c13c5dd7c13a64681a86afae63a5f347908daf103ce6d2f67
# via torch
kiwisolver==1.5.1 ; platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:072bdb15a3c19a5b5dbc8f8fb1f4e1884bf4f3507eeb4cc6334401274d37a5c0 \
--hash=sha256:34633ecf50d16187ab8e5528b7a2530f2feb4e23f300db4672538b51cfc5cd38 \
--hash=sha256:5daa1f19e097050b9c4d9a78fcc9263cb96c9dfae08037ddc1b7c4ad1889f2a2 \
--hash=sha256:74ad5c3dad54a4641b4c28cd15ded70899d04459c6c7aeacafea716be97cce6d \
--hash=sha256:95a02752aa032eef4aed01cda6d9b687c669bd0396bf4519eef8bba22a286720 \
--hash=sha256:aefe930d113798330e9462f7874542977869c0613cba3262e2de3a8d5dee8f3a \
--hash=sha256:b03af77d77e50edba2030fd5f7c352ff209314b09030a3cba7c14edf9a09a444 \
--hash=sha256:d09037ca068d784ebc4aec290ef952ca27ac15dd9c0b5801a88c6e1096b83e6b \
--hash=sha256:e12dfea7f5fc2a34a9080efbf79c4c44eb380ec5b9c6fea09407e08f0d1e941d \
--hash=sha256:f1303ef2eec81262a4b708c3e858afe58d7c75ad91c1c05266eda7673369859a
# via matplotlib
markdown==3.11.2 ; platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:87bf5d90f2d2882e2c02fcda182c2800456b985bff38913867c49706481514b0 \
--hash=sha256:f80390747a2a0ff2ba54da074ce4f862f72a6685f608cd2637b646a6d83993f5
# via tensorboard
markdown-it-py==4.2.0 ; platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:04a21681d6fbb623de53f6f364d352309d4094dd4194040a10fd51833e418d49 \
--hash=sha256:9f7ebbcd14fe59494226453aed97c1070d83f8d24b6fc3a3bcf9a38092641c4a
# via rich
markupsafe==3.0.4 ; platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:2628d3a8cb648ecebb3c5d6b0a1052d400e4d8b7ac0fb786be8d285b50040d17 \
--hash=sha256:2e9ad7dd851bf45fab9f75cbff4cb493fee9979e8d8c7c9c3ee119022518edd6 \
--hash=sha256:434139499bb20b502ed3baa1f169e618f924a97e7a777fea1a49446d80106cf6 \
--hash=sha256:50b5bedc9ed8a94fc8857a42ef4f84a81ea88f8d4f05dc8705fb23ee6d8dcca7 \
--hash=sha256:6da83a088f8ef93b2d483a8232a4dbf4d69d3d8496b568a03c56becac43e1808 \
--hash=sha256:7a83aa6e4805df46fed18e989d3d16f86ef60cb50bbc8d9ce3a6be89165fbf6e \
--hash=sha256:8e124f974786f831d6043728e38296969d3579db8896fe004682f5758e613581 \
--hash=sha256:c9a7f43c0b202b334cc9184af09bb8f21d3a209e038efaf106936fb69e6b026e
# via
# jinja2
# werkzeug
matplotlib==3.11.2 ; platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:07d9b9fa60cd4c393692f50d0bb03123242ddf61c99bb0e95e75feb354e7c1a8 \
--hash=sha256:116cdb0eb0eb5644fc98eb2975d4b4dd4ad35e5c8e6b851c22e6976f371f7ab5 \
--hash=sha256:1944895967f87c84c9b4bad29a707b31f5b36df4a3a2339ea1f8ea3ce5105539 \
--hash=sha256:3aa4b8516fd26659e4363abbf317c703d9116496c5db2e9d0609a2866dd39dd2 \
--hash=sha256:7e5a90f8a707ebb6004a713b1cff091a40cc5df4c7c1cb165e5a505ebc11c292 \
--hash=sha256:7ef7a53b66780e5d942923724f08577fbc5be1f7322da0f0f3f9dcaa45dd803d \
--hash=sha256:cec596316640f2b394b8f0daa0ea61a8eae82d017b620b9f202befb972a59ea4 \
--hash=sha256:df4f7784aca81a94f254c0a2767d592ee25f407e488f5fa7203e51093fb6ca27
# via mediapy
matplotlib-inline==0.2.2 ; platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:3c821cf1c209f59fb2d2d64abbf5b23b67bcb2210d663f9918dd851c6da1fcf6 \
--hash=sha256:72f3fe8fce36b70d4a5b612f899090cd0401deddc4ea90e1572b9f4bfb058c79
# via ipython
mdurl==0.1.2 ; platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:84008a41e51615a49fc9966191ff91509e3c40b939176e643fd50a5c2196b8f8 \
--hash=sha256:bb413d29f5eea38f31dd4754dd7377d4465116fb207585f97bf925588687c1ba
# via markdown-it-py
mediapy==1.2.7 ; platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:fc81a9190d9724b02286efac4e7a67f384024e078fda0a3cac9fdf25e2f8878f \
--hash=sha256:fd3826ca156f9165cd369f44d4d345d5609df1da3caf867df7ce304e19dd7fde
# via universal-humanoid
mjviser==0.0.16 ; platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:f50ea07ec4f35749eb9fd0470e921bc5ce6b8ae449926888c35fe8800ef3c8a0 \
--hash=sha256:fe8042eb1e97edbb310ac46e72e61d986b4dbe92048ef3c7a4598d2efe60ca53
# via universal-humanoid
ml-dtypes==0.6.0 ; platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:26b1f1fa4f0435a2946859823f6e2bf06796f1e9f10f5a05b08a5e3c8f46ff69 \
--hash=sha256:3b4a480aa8fd54a1805b8ac10f3f91763926a74f73c0c364c10f9231854f4170 \
--hash=sha256:5e60251d32ced5598972e4d5e06a2f044341f9291402551a3f6f0ec44f9299b0 \
--hash=sha256:6c8e39b53e90afda8ce52859c93de4dba3e02b76d85dcf091cc469f9184c6dae
# via
# onnx
# onnx-ir
# onnxscript
mpmath==1.3.0 ; platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:7a28eb2a9774d00c7bc92411c19a89209d5da7c4c9a9e227be8330a23a25b91f \
--hash=sha256:a0b2b9fe80bbcd81a6647ff13108738cfb482d481d826cc0e02f5b35e5c88d2c
# via sympy
msgspec==0.22.0 ; platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:0a13624a4969159fe35d8c2a3d377b2b61bbd8585e327440d5e52725affcce38 \
--hash=sha256:0b25dcbc108783cb72503ed705b9fbb8c3cb02ee5801923f44b5f038c91cc365 \
--hash=sha256:6f48317f05312bfdf78248f53933f830f07ab75cc1c813ac3ca4220cb3b5b019 \
--hash=sha256:99c401861c5bb3a57f7d6423ea7ed4352cd57aa3f04f4fbe9f3e3e4564a10f08 \
--hash=sha256:9a696f23f7c1ffb31fae308502e01a3965c3891d5c400f01d0d1096dbe77519e \
--hash=sha256:a6db3806b3b76ca78064255eac6fa101a8a64fe6f698d80fbaf81fdfa21217d4 \
--hash=sha256:c6c310ef83e7e291b01a63298828f848348bb99e84a1098c4b3923c05674d032
# via viser
mujoco==3.16.0 ; platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:76a5e87f89c68e6616519914cd8ac06dc0b60cd9108ef4aef92e3e506f59dc53 \
--hash=sha256:9990aecfe442a516d0595e5762f89da83431ed46fd353bee1e135e3a7140a551 \
--hash=sha256:a9ea38656fc75d9a6575b5fa7350882dc39461cb655e4ae8027ec58fda6245e1 \
--hash=sha256:bb8cfcf69cc0f3f2d07d988a3d8ac554bfea9b4a2331d9448a4c66c20722bfe4
# via
# mjviser
# mujoco-warp
# universal-humanoid
mujoco-warp==3.16.0 ; platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:00708e267fc2b2723c151fa4acf9f96f39765e43b845c28f5c81fb32eb534097 \
--hash=sha256:0852a1facaa97ddc9697e9818cf0e16cf4bbc893f95dab7b61b01882b1c79ee0
# via universal-humanoid
networkx==3.6.1 ; python_full_version < '3.12' and platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:26b7c357accc0c8cde558ad486283728b65b6a95d85ee1cd66bafab4c8168509 \
--hash=sha256:d47fbf302e7d9cbbb9e2555a0d267983d2aa476bac30e90dfbe5669bd57f3762
# via torch
networkx==3.7 ; python_full_version >= '3.12' and platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:e3fd2c13a7814cee3746340d8d7f8598a67f16a58bf47fb7f8793fab6efca1b0 \
--hash=sha256:fd77a511bd90f39f3d016351345b52cf5319b813bdca01de3f755d3cca62e96a
# via torch
numpy==2.4.6 ; python_full_version < '3.12' and platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:89cd468399cfd2504718f0ba50e410dca55a170b61a02ad92bb18c8a65186e93 \
--hash=sha256:8ad03c0965fb3c692200e74d458ca28c1dbb4ce96f9a479a8aa041ad5fabca02 \
--hash=sha256:90f9849678c75fe7afa2d348ac842c168b0a4d3d61919687216dfc547976d853 \
--hash=sha256:a0df0043bdb289bde1f62da130d20df23d58b45429f752bc7a8fc5325a225ecd \
--hash=sha256:a7830bab239b79cda9c08c2da014761cafb48da6150e1da17ac06283f43b6089 \
--hash=sha256:dfa20cc6ca228e6b155b11da03825975ce66aea520985dbbddf0f2a5a495c605 \
--hash=sha256:e5805d5a22fd19c8ccff10a9561f9df94436b0545619ea579db2d3c35294bce2 \
--hash=sha256:ede83e07a75dd06bc501566c1eca2afc0d61677c1472ac9ad93fdee6e638a48d \
--hash=sha256:f3a3570c4a2a16746ac2c31a7c7c7b0c186b95ce902e33db6f28094ed7387dda \
--hash=sha256:f407cb6b8e9d6d8c626bc73c945db1706035af8fd632295547bf1c9e46d092d6
# via
# contourpy
# imageio
# matplotlib
# mediapy
# mjviser
# ml-dtypes
# mujoco
# mujoco-warp
# onnx
# onnx-ir
# onnxruntime
# onnxscript
# scipy
# tensorboard
# tensordict
# torchrunx
# trimesh
# universal-humanoid
# viser
# warp-lang
numpy==2.5.4 ; python_full_version >= '3.12' and platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:6ffa07666f8da0eef81d149934a626d0d95fbd6838432a33e66245423a9062c0 \
--hash=sha256:9a94cf751c9ad8ebaa835bcd3d40dacf8534ad086b88c38029b65123c7999d2a \
--hash=sha256:b86966fbe4ad7de710422175572bcdc75fdedadfb54bc6fab7deabccddd7780b \
--hash=sha256:b89d0aaae2fe498c648f4c4795c084db535af5bd98ef942b2a3681fb74ce8645 \
--hash=sha256:fbde6962867ee75b48b0ee29b2b9372ec5d617799dbaf38e82dc0596f2f7738a
# via
# contourpy
# imageio
# matplotlib
# mediapy
# mjviser
# ml-dtypes
# mujoco
# mujoco-warp
# onnx
# onnx-ir
# onnxruntime
# onnxscript
# scipy
# tensorboard
# tensordict
# torchrunx
# trimesh
# universal-humanoid
# viser
# warp-lang
nvidia-cublas==13.1.1.3 ; platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:37936a16db8fe4ac1f065c2139360608a543a09275cb1a1af612e08cfa065436
# via
# cuda-toolkit
# nvidia-cudnn-cu13
# nvidia-cusolver
nvidia-cuda-cupti==13.0.85 ; platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:4eb01c08e859bf924d222250d2e8f8b8ff6d3db4721288cf35d14252a4d933c8
# via cuda-toolkit
nvidia-cuda-nvrtc==13.0.88 ; platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:ad9b6d2ead2435f11cbb6868809d2adeeee302e9bb94bcf0539c7a40d80e8575
# via
# cuda-toolkit
# nvidia-cublas
nvidia-cuda-runtime==13.0.96 ; platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:7f82250d7782aa23b6cfe765ecc7db554bd3c2870c43f3d1821f1d18aebf0548
# via cuda-toolkit
nvidia-cudnn-cu13==9.24.0.43 ; platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:71f181cd810e90f9b6023b01186fe82d13d65f0ec098581ee201d39fad769e4b
# via torch
nvidia-cufft==12.0.0.61 ; platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:6c44f692dce8fd5ffd3e3df134b6cdb9c2f72d99cf40b62c32dde45eea9ddad3
# via cuda-toolkit
nvidia-cufile==1.15.1.6 ; platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:08a3ecefae5a01c7f5117351c64f17c7c62efa5fffdbe24fc7d298da19cd0b44
# via cuda-toolkit
nvidia-curand==10.4.0.35 ; platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:1aee33a5da6e1db083fe2b90082def8915f30f3248d5896bcec36a579d941bfc
# via cuda-toolkit
nvidia-cusolver==12.0.4.66 ; platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:0a759da5dea5c0ea10fd307de75cdeb59e7ea4fcb8add0924859b944babf1112
# via cuda-toolkit
nvidia-cusparse==12.6.3.3 ; platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:2b3c89c88d01ee0e477cb7f82ef60a11a4bcd57b6b87c33f789350b59759360b
# via
# cuda-toolkit
# nvidia-cusolver
nvidia-cusparselt-cu13==0.8.1 ; platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:786ce87568c303fadb5afcc7102d454cd3040d75f6f8626f5db460d1871f4dd0
# via torch
nvidia-nccl-cu13==2.30.7 ; platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:cefa7fdb9710efd0f39c5f1be1d61ff6fc9a996c451265bd7fbdcf9455ed4b50
# via torch
nvidia-nvjitlink==13.4.92 ; platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:e0391f24ed94ec879b84e3da4d4ec320c879aff681f2c7a638462f7199284323
# via
# cuda-toolkit
# nvidia-cufft
# nvidia-cusolver
# nvidia-cusparse
nvidia-nvshmem-cu13==3.4.5 ; platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:290f0a2ee94c9f3687a02502f3b9299a9f9fe826e6d0287ee18482e78d495b80
# via torch
nvidia-nvtx==13.0.85 ; platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:4936d1d6780fbe68db454f5e72a42ff64d1fd6397df9f363ae786930fd5c1cd4
# via cuda-toolkit
onnx==1.23.2 ; platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:008cb0467b2bbee41448acc7da8b6f4e704624cb0d327a2d5adafc7ce19bc5b8 \
--hash=sha256:7abf381d278f31ac62487fddedc9dd42da842dce94d5d43536836ee3efdf4a2b \
--hash=sha256:c03ecf6b835d136108eeaeeafbd0026fc7b3cf98661409fbc6b63d5a29361348 \
--hash=sha256:e79e35e152d3095c6910ae81013bbc68679e32bfc0ca76f840968d4b6fdfb864
# via
# onnx-ir
# onnxscript
# universal-humanoid
onnx-ir==1.1.0 ; platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:76cc313a788c073b75271c1b6abb172aaf90fcc6335ed40f9f3238b2064a2fa3 \
--hash=sha256:b72176de575d820ec26864d17270d58ef886a7cf64db7649d15fcc3d5ae2a8ee
# via onnxscript
onnxruntime==1.31.1 ; platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:1f3dea47f2f0e5117c2404cea7d21fd253ae7836a3bf167a501cdfbb2600d536 \
--hash=sha256:5752bd508126ae7c49894d3b9d2e83f0c3d487319246d04c444b393308d7ad7f \
--hash=sha256:6711108bcddf8b401145fb31ac93776b8ca6e7e7e56acdfa9d478575b8df9a2f \
--hash=sha256:74a6a52994616c4d18b33db1e13c226333a70a6ba07480a5cab9c2a50dff1430
# via universal-humanoid
onnxscript==0.7.2 ; platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:2c664f6383d10f332a4d47b2876dcab16dba84909fe703656b19abc281fda165 \
--hash=sha256:d0e7121c6a1eefd608058928e111cbdb76709f70d269ff0d07aee493bd1d13c9
# via universal-humanoid
opentelemetry-api==1.45.2 ; platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:2cbd737377e710ee740d261c94e93fa0f08e6d037c18f17513cc0a2d94479a34 \
--hash=sha256:d7b0008e21e5fd61cbe7a7f34ca3b6905a78ab677d9bc395e299359c75cda6b4
# via
# opentelemetry-exporter-http-transport
# opentelemetry-exporter-otlp-proto-http
# opentelemetry-sdk
# opentelemetry-semantic-conventions
# wandb
opentelemetry-exporter-http-transport==0.66b2 ; platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:66d6ed810a515432c8891da9a834464332fd70badeca45bdeeae81b4280c903b \
--hash=sha256:fd5f5b490c099acda1cd7b909d5e8981436bf18cc8c6098d302d114c271720e9
# via opentelemetry-exporter-otlp-proto-http
opentelemetry-exporter-otlp-common==0.66b2 ; platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:06f5a9b88638e310efa1f1d206cdf6b75393e112aba4e174a963416192bd1028 \
--hash=sha256:ace86efb75489f925f1ef809856ae5f0c7db534e4afdc624530222228b1236a5
# via opentelemetry-exporter-otlp-proto-http
opentelemetry-exporter-otlp-proto-common==1.45.2 ; platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:af483b24217028892f017096731b794b196b785025d2e6808874720fd0b05a80 \
--hash=sha256:f1f2806f3f28ce7938033823e50934fcd91595a39de38e69a28a51e19f40fc6d
# via opentelemetry-exporter-otlp-proto-http
opentelemetry-exporter-otlp-proto-http==1.45.2 ; platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:5984a3fb46bc4628e4e2b85d4a6e72637e03962aae6cf372791137768a4ad939 \
--hash=sha256:6e4d025138829d0ee57946930c2ccad6491ad4d92a958e9c7eea69278fe60cfc
# via wandb
opentelemetry-proto==1.45.2 ; platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:03a7ca1b43f89fd5b965d5975902bb2515c0c86e2428b456836e9a6149c6235a \
--hash=sha256:bd8d23cf770605b0af9b7be2071d26a464fea289d35429abfd2a66226114ad64
# via
# opentelemetry-exporter-otlp-proto-common
# opentelemetry-exporter-otlp-proto-http
opentelemetry-sdk==1.45.2 ; platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:7eec4de13dd21ff3a1670dcb0e02079e692736f94d733dfdce5af461d884248d \
--hash=sha256:ff974ea70ee4f1d51e2ec152db37c2b626c7bf82bb3f7d86114fd8704e7199c6
# via
# opentelemetry-exporter-otlp-common
# opentelemetry-exporter-otlp-proto-http
# wandb
opentelemetry-semantic-conventions==0.66b2 ; platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:4cbe61e5f9275d57ab31708446e7938ef4291ebbe6a7fca6f7a629c5acae87d8 \
--hash=sha256:7da5f1a5dc4383eb66711d4d3d3e71824d9403ac1559fea021cc103c1a4c4976
# via opentelemetry-sdk
packaging==26.3 ; platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:94edc256424af38762eb31306eed28beb9f0efc50a8837492c9d6fd6004aed79 \
--hash=sha256:d7193f7c8e4e93f444fde0262bf90af30e16fa0ad0ad44cb553c87339b23cd1c
# via
# build
# matplotlib
# onnx-ir
# onnxruntime
# onnxscript
# pytest
# tensorboard
# wandb
paramiko==5.1.0 ; platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:239b5f9b2a53006cd54f83cf8495ee742083708ead947882c8dd2786de79f39b \
--hash=sha256:d9fe6d67a9d0bbf3eb85b051d26ab850b678c02d060f32915b944e5ca6d0d9e1
# via fabric
parso==0.8.7 ; platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:a8926eb2a1b915486941fdbd31e86a4baf88fe8c210f25f2f35ecec5b574ca1c \
--hash=sha256:eaaac4c9fdd5e9e8852dc778d2d7405897ec510f2a298071453e5e3a07914bb1
# via jedi
pexpect==4.9.0 ; platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:7236d1e080e4936be2dc3e326cec0af72acf9212a7e1d060210e70a47e253523 \
--hash=sha256:ee7d41123f3c9911050ea2c2dac107568dc43b2d3b0c7557a33212c398ead30f
# via ipython
pillow==13.0.0 ; platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:08b826a9bf1265c76de04a42061d61ea6b109418b7826f28ff1ecdfaa26c5502 \
--hash=sha256:23edf4ec4b6f4f240e8df6e4985e9ee2ced86474632433e55fb1e2548744bf40 \
--hash=sha256:2d50500a4a3bc6469adcdb319cf05d688b08ca3bf0f22a05339f3b25b433a599 \
--hash=sha256:42d63be3af4f9b951b199b0a9d23365d37ab14687e6d77b74940432c8515c4ca \
--hash=sha256:5f56499c1a74d2b862beb0c33b8016f99d095858455df9d21583bbbc1c15b582 \
--hash=sha256:83b37abd87b590acd8f47ca8a61f70893322b11834d36e07c2298dacad4d14d2 \
--hash=sha256:a625df9a2a5597fa5e5266e01aefc82604bb85d291e3547981e863560d75da6c \
--hash=sha256:cdcd1ba258178f7356d74170c9a81ba4eab6a5d55748bc402a5a2a6fe6be3510 \
--hash=sha256:e65dd6e75de67b2362a4fa3e16aa627595f65054dc4ec73d64680c84d592fa61 \
--hash=sha256:f391dc92a015c478745192fe4f2e120c2af2aba18d9fe0616341cbf537599806
# via
# imageio
# matplotlib
# mediapy
# mjviser
# tensorboard
platformdirs==4.14.3 ; platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:021add1ab113483edbcd4694f5fbb2d688b893ea30d35e1155bb371cf4be0be8 \
--hash=sha256:cba2d05fe5e18b26b1fa6f376b8e6234c012abe1f8696c486fface7d0054e931
# via wandb
pluggy==1.7.0 ; platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:7dd7b0d8832ba3cb632c306926ded123429211b83641b35dc5c41ad2d34f9bec \
--hash=sha256:d1eaa46ebb595891b860ab086b4d09c8588af65ebd4361b8e8f4bb8920b90ba8
# via pytest
prettytable==3.18.0 ; platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:439217116152244369caf3d9f1caf2f9fe29b03bd79e88d2928c8e718c95d680 \
--hash=sha256:b3346e0e6f79180833aebaac088ae926340586cf6d7d991b9eb125b65f72313a
# via universal-humanoid
prompt-toolkit==3.0.53 ; platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:01c0891d7f9237d5e339f7d3e42cdae80b7534abb1c7c0e3352efba6231492f2 \
--hash=sha256:9ec8a0ad96d5c56148b3f914aa79c1564c3fde5d2e6b876e7bc327e353cf8fa6
# via ipython
protobuf==7.36.2 ; platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:497d0463ff3316681da6c0b9e8d06cb465d61abce00b613ab42226175644d1bb \
--hash=sha256:89f23aa53c24553a2416fd4fd1ec06f74fa42b14b546d8883128813f775bbfd2 \
--hash=sha256:bdb3a345d48db958e6ce1f18e508beb0cc981d64f24088427549c866cd039f1e
# via
# googleapis-common-protos
# onnx
# onnxruntime
# opentelemetry-proto
# tensorboard
# wandb
psutil==7.2.2 ; platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:0746f5f8d406af344fd547f1c8daa5f5c33dbc293bb8d6a16d80b4bb88f59372 \
--hash=sha256:076a2d2f923fd4821644f5ba89f059523da90dc9014e85f8e45a5774ca5bc6f9 \
--hash=sha256:1a571f2330c966c62aeda00dd24620425d4b0cc86881c89861fbc04549e5dc63 \
--hash=sha256:b58fabe35e80b264a4e3bb23e6b96f9e45a3df7fb7eed419ac0e5947c61e47cc
# via ipython
ptyprocess==0.7.0 ; platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:4b41f3967fce3af57cc7e94b888626c18bf37a083e3651ca8feeb66d492fef35 \
--hash=sha256:5c5d0a3b48ceee0b48485e0c26037c0acd7d29765ca3fbb5cb3831d347423220
# via pexpect
pure-eval==0.2.4 ; platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:260c2774686e651b79f8b8e7fc9d80b3599ea6a66334b47d5f4abb69fc2c0ea1 \
--hash=sha256:96cae060a313cfaad51bb761278bfb0e62dc0248d9315a81173752dc546cd37a
# via stack-data
pycparser==3.12 ; implementation_name != 'PyPy' and platform_machine == 'x86_64' and platform_python_implementation != 'PyPy' and sys_platform == 'linux' \
--hash=sha256:a076f3f489b39b43ed13563a1dafe42447445bec2e294154ec5c57fcdda4c85e \
--hash=sha256:f03387877eab82ce01bbdfd0ba7252ee55a90b1c708d5129e2ce915475cf1db6
# via cffi
pydantic==2.14.2 ; platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:7efbf7507b937931cd32a5755acb00ea1cf5ccb1d4e8067a5850d5069e382118 \
--hash=sha256:fb7b2db6cefe8414dd612edeede49c4c433acb0432b0f92b95024e45f66c505d
# via
# cwevals
# wandb
pydantic-core==2.50.2 ; platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:16ec8e07c37f02dcf68939cc7f202ba9a148cc155ee2acf22d337e10b24c2cb7 \
--hash=sha256:1b0e8dc3932cf4e9ed5727570325cacc716b9409ffc794a221c0ccc72e8eb775 \
--hash=sha256:31a1fb3ffbd1195b24490f671d12623dd4c9bfa56d6af0a299fa43b516134f5b \
--hash=sha256:86e7abd24a67f61a4bfe8df059b67a2bdd8bf33705178e98463195d140bd3780 \
--hash=sha256:8772c2c0091e6fc2d0499151793d31bf811826986b0727052935a95104845544 \
--hash=sha256:91238117cd296352f54f24117ff5a227bdea428865dba1244aa0cdd1424c0bc2 \
--hash=sha256:928a44ad1e21a7e923eb924d601605f0694cd187d2ce5d40bf7ad76650f4627d \
--hash=sha256:92ada606285dab054e40848a951f583c537cde1e54701ea1305ec0096877012d \
--hash=sha256:9667637665acba166a7dbbc786f14c50512d3a01f7199e0e1a13b929e92d7c7a \
--hash=sha256:9e7460abce637f90e37f8c443559fbd687c38884d05cb69ef265e634945f4324 \
--hash=sha256:b5675a12a7b67063fe3ca187e735b0463cf13fabc10166ae4d4b2f775fd0632d \
--hash=sha256:cb8d2d7b07cbe2d7488de7e53671f836c4f4a8032dcda079df33dc1b7e2397c0
# via pydantic
pygments==2.21.0 ; platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:2363c69b61c4a97c838da3b130dcd6468f4848992b21a82f2a63ec34377137d9 \
--hash=sha256:610ca751c9bc2492b38eb9a38a7fbc93edbbb2d7182edaf34e66ae493dee5c8c
# via
# ipython
# ipython-pygments-lexers
# pytest
# rich
pynacl==1.6.2 ; platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:018494d6d696ae03c7e656e5e74cdfd8ea1326962cc401bcf018f1ed8436811c \
--hash=sha256:22de65bb9010a725b0dac248f353bb072969c94fa8d6b1f34b87d7953cf7bbe4 \
--hash=sha256:8a66d6fb6ae7661c58995f9c6435bda2b1e68b54b598a6a10247bfcdadac996c \
--hash=sha256:8b097553b380236d51ed11356c953bf8ce36a29a3e596e934ecabe76c985a577 \
--hash=sha256:c8a231e36ec2cab018c4ad4358c386e36eede0319a0c41fed24f840b1dac59f6
# via paramiko
pyopengl==3.1.10 ; platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:794a943daced39300879e4e47bd94525280685f42dbb5a998d336cfff151d74f \
--hash=sha256:c4a02d6866b54eb119c8e9b3fb04fa835a95ab802dd96607ab4cdb0012df8335
# via mujoco
pyparsing==3.3.3 ; platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:928ae7e20211f3b6f3915a72f06a0cfd29ab9d24279dd6346b6b1a7146397d36 \
--hash=sha256:ece8c00a69cf01b45d0b1dedabb469c90d8caf996d4fda40f147627a122849a4
# via matplotlib
pyproject-hooks==1.3.3 ; platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:5fc53fdac9f7bd63fbcdc868fb5f90b4784d78a53a3d3388cd738b807441a20b \
--hash=sha256:defda19b854fa0d3bd4f76ea4ddcba8abd7dcfcdd585a6690ade050744fc5f43
# via build
pytest==9.2.0 ; platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:07acdcf88ec174f907dc01af70bf52d49b014bbab03faeeade31c96b18c2e588 \
--hash=sha256:97c4f3b1f3c7414aa36465e09b8220f1d03627993bd51b1282e193753c80557b
# via universal-humanoid
python-dateutil==4.0.0 ; platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:a5e246f79eb80475e691bf93862cead98decd8a498cf53ba0516fcdc0f50dc1c \
--hash=sha256:d7295ff14fdc0f92f662100bfb6c551d160ca090f2939f1ec5b9794265c03f06
# via matplotlib
pyvers==0.2.3 ; platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:6f5b5612f2f4bd08caa49baa70fc5f875fc7da701a5385c13e244eea6b8114dd \
--hash=sha256:c4b81c3a033963245e124cdecb052783c9c4cea3bb08c051833af1c44faa6283
# via tensordict
pyyaml==6.0.3 ; platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:0f29edc409a6392443abf94b9cf89ce99889a1dd5376d94316ae5145dfedd5d6 \
--hash=sha256:37503bfbfc9d2c40b344d06b2199cf0e96e97957ab1c1b546fd4f87e53e5d3e4 \
--hash=sha256:41715c910c881bc081f1e8872880d3c650acf13dfa8214bad49ed4cede7c34ea \
--hash=sha256:b8bb0864c5a28024fac8a632c443c87c5aa6f215c0b126c449ae1a150412f31d \
--hash=sha256:ba1cc08a7ccde2d2ec775841541641e4548226580ab850948cbfda66a1befcdc \
--hash=sha256:d76623373421df22fb4cf8817020cbb7ef15c725b9d5e45f17e189bfc384190f \
--hash=sha256:eda16858a3cab07b80edaf74336ece1f986ba330fdb8ee0d6c0d68fe82bc96be
# via
# universal-humanoid
# wandb
requests==2.34.2 ; platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:2a0d60c172f83ac6ab31e4554906c0f3b3588d37b5cb939b1c061f4907e278e0 \
--hash=sha256:f288924cae4e29463698d6d60bc6a4da69c89185ad1e0bcc4104f584e960b9ed
# via
# opentelemetry-exporter-http-transport
# opentelemetry-exporter-otlp-proto-http
# viser
# wandb
rich==14.3.4 ; platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:07e7adb4690f68864777b1450859253bed81a99a31ac321ac1817b2313558952 \
--hash=sha256:817e02727f2b25b40ef56f5aa2217f400c8489f79ca8f46ea2b70dd5e14558a9
# via viser
ruff==0.17.3 ; platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:692fdb46d0fd9c71e380947dc5fc72f2c3f881073bc1f739e4579376231affbe \
--hash=sha256:7fa4b0f931cdbcb8a52044f370a3eb429fb42eb88807b0e800fcaa98dd42b24e \
--hash=sha256:a57049c8addb3662073859bb55a7ad35aa54f8a7c9ed268911d1333c5aadaaa9
# via universal-humanoid
scipy==1.17.1 ; python_full_version < '3.12' and platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:02ae3b274fde71c5e92ac4d54bc06c42d80e399fec704383dcd99b301df37458 \
--hash=sha256:43af8d1f3bea642559019edfe64e9b11192a8978efbd1539d7bc2aaa23d92de4 \
--hash=sha256:4eb6c25dd62ee8d5edf68a8e1c171dd71c292fdae95d8aeb3dd7d7de4c364082 \
--hash=sha256:581b2264fc0aa555f3f435a5944da7504ea3a065d7029ad60e7c3d1ae09c5464 \
--hash=sha256:6609bc224e9568f65064cfa72edc0f24ee6655b47575954ec6339534b2798369 \
--hash=sha256:74cbb80d93260fe2ffa334efa24cb8f2f0f622a9b9febf8b483c0b865bfb3475 \
--hash=sha256:94055a11dfebe37c656e70317e1996dc197e1a15bbcc351bcdd4610e128fe1ca \
--hash=sha256:95d8e012d8cb8816c226aef832200b1d45109ed4464303e997c5b13122b297c0 \
--hash=sha256:f590cd684941912d10becc07325a3eeb77886fe981415660d9265c4c418d0bea
# via universal-humanoid
scipy==1.18.1 ; python_full_version >= '3.12' and platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:52c4b7422442aba924d03ad4019852b08a92e64ea187b933135687bfe2747307 \
--hash=sha256:d2924a03db38dc2e848bca2fe9f077dafb891480b91a00a0963a8cf86dfc31c1 \
--hash=sha256:e3b417bf8c2c7c16e8f58ad91db17783ec911ac16e7b50eb6eab6e809b4f5b07 \
--hash=sha256:f55fa87b6c612ecd6b058f167c53231b1d14e412efe361d3d6e38b3631c73218 \
--hash=sha256:fdaf5ea890a6183d0565f51a61799d67081bd5b1cf03c5f4b3fd3732108625c9
# via universal-humanoid
setuptools==82.0.1 ; platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:7d872682c5d01cfde07da7bccc7b65469d3dca203318515ada1de5eda35efbf9 \
--hash=sha256:a59e362652f08dcd477c78bb6e7bd9d80a7995bc73ce773050228a348ce2e5bb
# via
# tensorboard
# torch
# universal-humanoid
smmap==5.0.3 ; platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:4d9debb8b99007ae47165abc08670bd74cb74b5227dda7f643eccc4e9eb5642c \
--hash=sha256:c106e05d5a61449cf6ba9a1e650227ecfb141590d2a98412103ff35d89fc7b2f
# via gitdb
sniffio==1.3.1 ; platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:2f6da418d1f1e0fddd844478f41680e794e6051915791a034ff65e5f100525a2 \
--hash=sha256:f4324edc670a0f49750a81b895f35c3adb843cca46f0530f79fc1babb23789dc
# via cwevals
stack-data==0.6.3 ; platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:836a778de4fec4dcd1dcd89ed8abff8a221f58308462e1c4aa2a3cf30148f0b9 \
--hash=sha256:d5558e0c25a4cb0853cddad3d77da9891a08cb85dd9f9f91b9f8cd66e511e695
# via ipython
sympy==1.14.0 ; platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:d3d3fe8df1e5a0b42f0e7bdf50541697dbe7d23746e894990c030e2b05e72517 \
--hash=sha256:e091cc3e99d2141a0ba2847328f5479b05d94a6635cb96148ccb3f34671bd8f5
# via
# onnx-ir
# torch
tensorboard==2.21.0 ; platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:7279316dcb6bd5bc391d623dea841531299cde1887310e8133bc34a996d32255
# via universal-humanoid
tensorboard-data-server==0.7.2 ; platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:7e0610d205889588983836ec05dc098e80f97b7e7bbff7e994ebb78f578d0ddb \
--hash=sha256:ef687163c24185ae9754ed5650eb5bc4d84ff257aabdc33f0cc6f74d8ba54530
# via tensorboard
tensordict==0.15.4 ; platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:e3bacbbb739b2c5aa535dc73c72a5cf6afd8843c1681225946f5aae5c5ffd5f2
# via universal-humanoid
torch==2.14.0+cu130 ; platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:07686af6e00801b04fe0feb9e9aa22ff7a58112fb60e0b29a4146f5ac553bf6b \
--hash=sha256:745010695e0458d6f28accb697b5371b6fd245aa8696530165f972cd126bbd8d \
--hash=sha256:9c5693d3ca91743202034094ad3b3df20b5eb1373d5e0bc8a5d976be12c35302
# via
# tensordict
# torchrunx
# universal-humanoid
torchrunx==0.4.0 ; platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:a8aa42292941ddb7d68fe42a5eabe25800176a12250d417192bd7501b839e52f \
--hash=sha256:adcdbfd82ef633e549ba6e7694bc6eb258f55a6a64c8c59776794b922ae7b3ff
# via universal-humanoid
tqdm==4.70.1 ; platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:c293e525e6fef9c20e8728fd4612df02a0aa31bb5fe91ecd93e123b1b7bffa73 \
--hash=sha256:cefd0eca11b2a37a3aee776544d4f4ae913f02688135b5556b8788dfa474afc4
# via
# universal-humanoid
# viser
traitlets==5.16.1 ; platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:ed900c2b631aa3a112811139fa97b8d2c3bad5e989656bba4b7e52c7852c18c1 \
--hash=sha256:f775618166caa0396c8e337099240f2bd3e5e917d203b2e6fbe21a58d3cb1f6b
# via
# ipython
# matplotlib-inline
trimesh==5.2.0 ; platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:04db2c7ec62afdd7763fb04c82bb5f257ddd984993a19ee83697f475fc58fd29 \
--hash=sha256:d0a845d058e5a4ff145d0f0082d80c58d761687ce389c79d422dbe774966957b
# via
# mjviser
# universal-humanoid
# viser
triton==3.8.0 ; platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:1f0497218e26b7d79773ad9c2a3fa3b539ee69f587a13fac2e552b1d322a8015 \
--hash=sha256:68988ac85d5e7086baeda0ddc175af9667db7529b3c5e11a5c0601b8bef2200a \
--hash=sha256:e91ffa46d095b252248297292dd22bcbacd53a125a0c2eefbbbf74925a320bc3
# via torch
typeguard==4.6.0 ; platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:79878165bb86f2cf5d41d159a0ff1792a796cf496882d2fe1b1c6c7049b9cdd7 \
--hash=sha256:e7414f09111317de3e335de92cd397c5c0ca00b1cc1676de12e1d444a79b3f21
# via tyro
typing-extensions==4.16.0 ; platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:481caa481374e813c1b176ada14e97f1f67a4539ce9cfeb3f350d78d6370c2e8 \
--hash=sha256:dc983d19a509c94dba722ee6abd33940f7c05a89e243c47e907eb4db6f1a43e5
# via
# anyio
# cwevals
# etils
# grpcio
# ipython
# onnx
# onnx-ir
# onnxscript
# opentelemetry-api
# opentelemetry-exporter-otlp-proto-http
# opentelemetry-sdk
# opentelemetry-semantic-conventions
# pydantic
# pydantic-core
# torch
# typeguard
# typing-inspection
# tyro
# viser
# wandb
typing-inspection==0.4.4 ; platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:547274fa6b0a561ccf549cc9524b999a578e737d015d8709d021f9d0d13bea47 \
--hash=sha256:65b8397ba37ccbce054456aaccddfc91e6e3083c92824df348d96ca832f3f147
# via pydantic
tyro==1.0.16 ; platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:2f9087196113699647190459a12b6b4f2784dd0703c50ab1f7bf0f85a3a794ea \
--hash=sha256:edf23c9c3eecba5a2b5af152df8f34a723b689a001ff5b7311f93eff1893a636
# via universal-humanoid
tzdata==2026.6 ; platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:0a10eec6e13ef16b627fabed9a5f2ba61a62b1c67f327305f4e7e412a70b4aec \
--hash=sha256:85d2cd9c09f81f0bb29db27759fc83af4f1e620efb814f8c5bf8b4cc4feece6d
# via python-dateutil
urllib3==2.8.0 ; platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:0cf3cae568d36aa9576b28dfb35f11328f1cb974ca7647d9475ebb86c75ac6e3 \
--hash=sha256:63bf2ead4c879426ebf22ef2a781eeb4aa3b4ae798a0435506f8687fd5bb9b63
# via requests
viser==1.1.1 ; platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:0a7ba97dab32b8a7106a2dd216fe1373f2e66e17529b5b43502d33d9b9ac9352 \
--hash=sha256:f528877553c56e1c01d43b861a1434920a2e7ba4fdc7b2e7ff03721e3c3f5b3d
# via
# mjviser
# universal-humanoid
wandb==0.31.0 ; platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:009bcf370598d5e7a3d3bb66e33510f78c17fcc0505f8aa7ecf51dbc4378f979 \
--hash=sha256:dcdaa635576ff17b18a757b39fc22b0db3d1c6e01d5ce090fe80f6adc776028d \
--hash=sha256:df16a46cd6468e6580783b02560c79991c181f63a4691949020fff1b4ac87072
# via universal-humanoid
warp-lang==1.18.0 ; platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:090962bfae7dd22ade0d8e0258f41635e0e58980f15bcc9e7a7d312819dfe440
# via
# mujoco-warp
# universal-humanoid
wcwidth==0.9.4 ; platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:901eace6d2dd25ebb917d0ce5db18e5c537e1504c905cf32f0d006c12098b05d \
--hash=sha256:b185c8b927b23af2721be1783499d1cfa5c028b098b0858951b4ae224477f1c2 \
--hash=sha256:d8905f1390d828d745fdd1904374c90386d902d3cfa9ee88e9875099f75a4e17 \
--hash=sha256:fd8d57d3cdfe570e3265c8fab1885c24e34ef40889c0759c0c304a19e9835954
# via
# prettytable
# prompt-toolkit
websockets==16.1.1 ; platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:0f62863e8a00a6d33c3d6566ec0b89f23787b747ffe0c3bc71ec0e76b82c94b1 \
--hash=sha256:30bbe120437b5648a77d3519b7024ea09530e0b5b18d3698c5a0ae536fe0cc2e \
--hash=sha256:53260c8930da5771cec89439bff99c20c8cb03ddb9588b980697355a83cd4bd3 \
--hash=sha256:6abbd3e82c731c8e531714466acd5d87b5e88ac3243465337ba71d68e23ae7e3 \
--hash=sha256:d14bfb217eb4701e850f1525c9d29d79c44794cdf1c299ead25f39f8c78dea81 \
--hash=sha256:da4ca1a9d72f9030b3146b8d7022719a9f3d478f61efe6f7dd51d243f61c51b2 \
--hash=sha256:db234eda965dcce15df96bb9709f587cd87d4d52aaf0e80e2f34ec04c7670c57 \
--hash=sha256:ed5bb271084b46530ee2ddc0410537a9961152c5ccba2fc98c5276d992ccba87 \
--hash=sha256:f5d497865f05bb222cab7016c6034542e84e5f29f49c6fd3f4939cda7197b5b8
# via
# mujoco
# viser
werkzeug==3.1.9 ; platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:55ca7c70a75689be937aa27f8ff4b018f06ff4838fc73045560bf0f5a1291060 \
--hash=sha256:6392e50c78460ba618e5b21f08a71f59c99ce99cdc6cf6e3dd7e6ccca8754fab
# via tensorboard
wrapt==2.5.2 ; platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:0743950d3dda9940e56f81f5334ba051c4fe2e1421e4e4a4ef8eb8fb86a90894 \
--hash=sha256:092472b7b1494deef40dac5e9f6295013a9d51b112b5fbc4350f64dc2574e46a \
--hash=sha256:40e03506f09dbf4854fd69848e843dadd9b7fbbdd217837cb27b37fb8f0b5bc3 \
--hash=sha256:551cb49e88b3722faad780303833bc4fa70685fb1773270fb98b40d39d9fddbf \
--hash=sha256:61eb54e9540ea522c2c4752a57549a83d3c67e6704ed708403d28e067aaf90f0 \
--hash=sha256:7a7efde469ad904b8e05f98c76d63b437f4756c30ec7b9136bef02bdb5d1b14c \
--hash=sha256:8b039b3a296762db1908d6108365225393c4fb33c1dbc5f00aa36f8302eaeaa0 \
--hash=sha256:eff91af619ad5d587f8855f6c02164e12d33d205c702eb73d487deb499520265
# via deprecated
xxhash==4.0.2 ; platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:050c1d66587e8f1ce70ccb7f661ec117e01497d48da817c14543beba0b8c677e \
--hash=sha256:180b349bbc89af61240adab8519426e0cb6546c2487e4b26b464364bf59dcbed \
--hash=sha256:42fcaec8c705d1ea362d4cf011aea0dad1ebe13096ed227f9cfecfcac396c72c \
--hash=sha256:56ee43c284daffc12d269b3fb36b1fc624c7ab007064a24d54099049ee71ce80 \
--hash=sha256:61ded7dee4f95da69295c228fc18a65447f5741c45a315aa3f9a3adda6038611 \
--hash=sha256:63522739ce94e4e73c1605aa60a52b3863e77e835f6292cc8a5e7e7b626ff00c \
--hash=sha256:79cfd336c2ecaceef1061e7a7b3da079eab5920ba1352a28be690e857f67f2ea \
--hash=sha256:7c7c95dba490be3b4e5a2fa5595b4d4f088fe7b8726e5998b8c1d83b1f9eb3e4 \
--hash=sha256:87255720aa1a1be35256b688c5ee4151d3273894186cda5c8a5b82f1dc05debf \
--hash=sha256:892a66641657cf9aa0ae5c2c88258bc281221a33bd4af34bed786eea23e06d83 \
--hash=sha256:af3d1b3cfa5280e84d8968e9f30c9eca75781db4915eb6d6acb0dd4fc93e0f47 \
--hash=sha256:c37fcacb7f5b8590c822bbf3bc41690e40a7581d8c4bcea7b15148f0743e716d \
--hash=sha256:e359ff2f38be3c856a129c1c5c4cbd70e219188456c1bf266fd4f94e2baae278 \
--hash=sha256:f8e86a0b5e28418162dfa68d6dc4bee873ea2efbd6ee67789903a0ee37a81c14
# via wandb
zipp==4.1.2 ; platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:b30cb6cd253122169c47de245ca24a86bac2a5d3802ba15bf7ba6414eb9e144d \
--hash=sha256:cf474b90b07ea932637e17c6b9479de2b2882900b0d311bdaf4f5c000a15a97f
# via etils
zstandard==0.25.0 ; platform_machine == 'x86_64' and sys_platform == 'linux' \
--hash=sha256:172de1f06947577d3a3005416977cce6168f2261284c02080e7ad0185faeced3 \
--hash=sha256:474d2596a2dbc241a556e965fb76002c1ce655445e4e3bf38e5477d413165ffa \
--hash=sha256:4f187a0bb61b35119d1926aee039524d1f93aaf38a9916b8c4b78ac8514a0aaf \
--hash=sha256:5a56ba0db2d244117ed744dfa8f6f5b366e14148e00de44723413b2f3938a902 \
--hash=sha256:72d35d7aa0bba323965da807a462b0966c91608ef3a48ba761678cb20ce5d8b7 \
--hash=sha256:7713e1179d162cf5c7906da876ec2ccb9c3a9dcbdffef0cc7f70c3667a205f0b \
--hash=sha256:8e735494da3db08694d26480f1493ad2cf86e99bdd53e8e9771b2752a5c0246a \
--hash=sha256:9300d02ea7c6506f00e627e287e0492a5eb0371ec1670ae852fefffa6164b072 \
--hash=sha256:aaf21ba8fb76d102b696781bddaa0954b782536446083ae3fdaa6f16b25a1c4b \
--hash=sha256:f373da2c1757bb7f1acaf09369cdc1d51d84131e50d5fa9863982fd626466313
# via viser
+395
View File
@@ -0,0 +1,395 @@
{
"added_since_import": [
"rl/mjlab/licenses/IsaacLab.txt",
"rl/mjlab/licenses/mujoco_menagerie-i2rt_yam.txt",
"rl/mjlab/licenses/mujoco_menagerie-unitree_g1.txt",
"rl/mjlab/licenses/mujoco_menagerie-unitree_go1.txt",
"rl/rsl_rl/licenses/pfrl.txt",
"rl/upstream/import.json"
],
"changed_since_import": [
"rl/mjlab/asset_zoo/robots/booster_t1/t1_constants.py",
"rl/mjlab/sim/sim.py",
"rl/mjlab/tasks/velocity/mdp/velocity_command.py"
],
"files": {
"rl/mjlab/LICENSE": "8ca04d9dbbccd20f0f1842fb33e13949f9e57064b150dad36af8e7a751ca3321",
"rl/mjlab/__init__.py": "7a447eb4e744ff67a0557c32c3aa1a27c352b7eaccc6df80ebb1b9cad71a646b",
"rl/mjlab/actuator/__init__.py": "ae89d8f9cab9ab238a881597ec94f729feee7058b0d9773dca83000783a790df",
"rl/mjlab/actuator/actuator.py": "85c5744fe149d7a7886450f15e30317f4b01a3d2f1807730946fdd808e91716e",
"rl/mjlab/actuator/builtin_actuator.py": "4bd44d70c59a150313a0e8321c4700d8ab85c856c2554d55507d5d99e3544c03",
"rl/mjlab/actuator/builtin_group.py": "f4429b846fc645c443734e45435618958d9f24502b2c627156e7d5a69cbbd9e3",
"rl/mjlab/actuator/dc_actuator.py": "a24fc574b1606d66338bde4905e83d24ca2e66a2f7b11e0235fa441a81d253f6",
"rl/mjlab/actuator/fused_group.py": "21a5d5ca68c0b886ba5d5abf89665bbc8c63991eb0bad9f4cb204f5300df46e1",
"rl/mjlab/actuator/learned_actuator.py": "aefb27cc12881e67c87171cb94e38a5aee0171299ffd952e07053fbbd7fac487",
"rl/mjlab/actuator/pd_actuator.py": "b2c36f1759827890055cb3549aa7b441a50abd914c883fde8bbfd631e6f2f101",
"rl/mjlab/actuator/xml_actuator.py": "a8df6bbf07dda9e182d5ef248b0022816d1aa211ba321fa00ab676286b9c9548",
"rl/mjlab/asset_zoo/README.md": "e6bed2af0c36ad4768c4cdba7710ff07354a76f5e0573ec60070baa69212b9db",
"rl/mjlab/asset_zoo/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
"rl/mjlab/asset_zoo/robots/__init__.py": "7b64b50e70acf42fca4be9fe23b4793c725876847a81bd0fa540a7e55e60e449",
"rl/mjlab/asset_zoo/robots/booster_t1/__init__.py": "6630c30df59a46d1f341f8dfc350cb99a723a839ec643745a1fa0790d60bfdf4",
"rl/mjlab/asset_zoo/robots/booster_t1/assets/LICENSE": "056db8cc7dcb21feae042d97185d4689c21bb1ba7fe59cefdc7ef5917c0ea77f",
"rl/mjlab/asset_zoo/robots/booster_t1/assets/T1_23dof.xml": "f00e3410362b1e2267e6db05d25c33d6240bbf2436d62a409f07042bad4f7fb2",
"rl/mjlab/asset_zoo/robots/booster_t1/assets/meshes/AL1.STL": "43f5d268152b3c169d34731545ddfa2a510a7e16f209a2a80f8ba9ae0e1a40aa",
"rl/mjlab/asset_zoo/robots/booster_t1/assets/meshes/AL2.STL": "4bca60adb29c73f701350abd23e8a8d660dc61e87c5163a80157650dd3ad88c3",
"rl/mjlab/asset_zoo/robots/booster_t1/assets/meshes/AL3.STL": "bfdf5611dfe90f8e02b07b037c36a244f4ce7184b45113371ef46e624fb30854",
"rl/mjlab/asset_zoo/robots/booster_t1/assets/meshes/AR1.STL": "6dbbd3dab4486600747be9d0e7f8523d67bf00d24d0dc9c23db1766c4d808fcf",
"rl/mjlab/asset_zoo/robots/booster_t1/assets/meshes/AR2.STL": "0499136785fe1bc0289d9e503e1e0ea48f5a63420a2e45b0fcd9e6b3bcf8f999",
"rl/mjlab/asset_zoo/robots/booster_t1/assets/meshes/AR3.STL": "92330bd8286fe420fb76949ee4c5d5bdec2e6bb30426acdc4d6fae9257464631",
"rl/mjlab/asset_zoo/robots/booster_t1/assets/meshes/Ankle_Cross_Left.STL": "e6ec38741838475834c147f2ad18196c0254741e602c2b676f888fd12b5d7981",
"rl/mjlab/asset_zoo/robots/booster_t1/assets/meshes/Ankle_Cross_Right.STL": "b15d91f0f2470fc4d5634db906a3d7efaf1f700302e6671824494da8faf50308",
"rl/mjlab/asset_zoo/robots/booster_t1/assets/meshes/H1.STL": "527a2e426db4e3d476a195754764b7f9565a950b809a34e47618f177b139c3f2",
"rl/mjlab/asset_zoo/robots/booster_t1/assets/meshes/H2.STL": "9cbf4fe83a721c809de65b1f0d0a379314f4523f7e2f8af7f0edc27c2ce385aa",
"rl/mjlab/asset_zoo/robots/booster_t1/assets/meshes/Hip_Pitch_Left.STL": "7bcb8be569aeb5379cb75a0827d682bc1d73a52a7f515d3aeffece43842c33c9",
"rl/mjlab/asset_zoo/robots/booster_t1/assets/meshes/Hip_Pitch_Right.STL": "054d579c3c6ccc1f97aaf1461d802dec748ce1bf26c6fdd289283a2d0dbb2032",
"rl/mjlab/asset_zoo/robots/booster_t1/assets/meshes/Hip_Roll_Left.STL": "72dbd65bdb4e87444c8e538474f631d63bf8c0960922517456b8cf2be9e1d0f0",
"rl/mjlab/asset_zoo/robots/booster_t1/assets/meshes/Hip_Roll_Right.STL": "d715a7649ea4bf20d7ee17b3a57bae31732e467b9da1053fee2242c90f50980f",
"rl/mjlab/asset_zoo/robots/booster_t1/assets/meshes/Hip_Yaw_Left.STL": "8d7de83ab4034639db3d1be1e6f43ae16e004dc2f1c2bec170b615686d187918",
"rl/mjlab/asset_zoo/robots/booster_t1/assets/meshes/Hip_Yaw_Right.STL": "1a7d98e021884e313c55780a3c7c24ef2aa322807d8f9f8b18b2a63c7e2b9bf6",
"rl/mjlab/asset_zoo/robots/booster_t1/assets/meshes/Logo.STL": "221b00d090fdc9e684892805db75265c9688b152b3ccf641084f950f220596c8",
"rl/mjlab/asset_zoo/robots/booster_t1/assets/meshes/Shank_Left.STL": "6ae8dd7f88cb70a4a3cb3972cad4035a76f1604fd73ea66e285da553b3703c40",
"rl/mjlab/asset_zoo/robots/booster_t1/assets/meshes/Shank_Right.STL": "dc66a8a0e2e31e413884a053e427d9b5007bef4508986587916a638a8a448599",
"rl/mjlab/asset_zoo/robots/booster_t1/assets/meshes/Trunk.STL": "39d9610b23ef8ac7ea4343231a499f9c485903bede62df9729a2cd1254fede8f",
"rl/mjlab/asset_zoo/robots/booster_t1/assets/meshes/Waist.STL": "2fe0a9ccfc284beca402e554db38e83022592192afa00f02db13cb374ffe9fbc",
"rl/mjlab/asset_zoo/robots/booster_t1/assets/meshes/left_foot_link.STL": "2de026f2269fc55f80d4a8abd2e33dfd152adf9eb2c240d7d7e1dc40c13b55aa",
"rl/mjlab/asset_zoo/robots/booster_t1/assets/meshes/left_hand_link.STL": "34df992aed438446481cfa20e78327810cd5cc5ecbb4a6f32ba2bf625462f53d",
"rl/mjlab/asset_zoo/robots/booster_t1/assets/meshes/right_foot_link.STL": "313a9ff2f95da5727dce2b98c50b60226bf0a3affc72050ee251ea503e4b370a",
"rl/mjlab/asset_zoo/robots/booster_t1/assets/meshes/right_hand_link.STL": "c5a73e493e17db7611a81785f80c01167f51018271b4042f72fa8d2351d8b0eb",
"rl/mjlab/asset_zoo/robots/booster_t1/t1_constants.py": "04c351bff19f6d63d860f5eebfb41eb783a616e58298254d140dadf9d68c2cc7",
"rl/mjlab/asset_zoo/robots/i2rt_yam/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
"rl/mjlab/asset_zoo/robots/i2rt_yam/xmls/assets/d405.stl": "6617f3fb22b604a66c27e16d2397f305fcdca8a781248abdbabbf69c9fe6a3c5",
"rl/mjlab/asset_zoo/robots/i2rt_yam/xmls/assets/model2.stl": "5ac61c0313ed00c38655e41681af1b8b307cf8d5ef57209a75f215cea02df754",
"rl/mjlab/asset_zoo/robots/i2rt_yam/xmls/assets/model2__10.stl": "b879cc07f547b7a2ac52e428dd206e80a0d4b0e805cb56b9a8230df04a690923",
"rl/mjlab/asset_zoo/robots/i2rt_yam/xmls/assets/model2__11.stl": "af220d2a51e8444392aef36e2fe4a6f69aed22bc292a7f38e54da3b4f5253434",
"rl/mjlab/asset_zoo/robots/i2rt_yam/xmls/assets/model2__12.stl": "e79614869bd2283f70fafa150ce748d5b6c03377a651448c6fe9893a22f6d1de",
"rl/mjlab/asset_zoo/robots/i2rt_yam/xmls/assets/model2__13.stl": "32b62f14f00ea82bfc79b935f7dac32c7cb276a3abc0aae8362cbba3fb1326de",
"rl/mjlab/asset_zoo/robots/i2rt_yam/xmls/assets/model2__14.stl": "72d2178789dd5e0358c2caf9dad90b8881e505f4c52a8d6d18c71b5e93e8c84b",
"rl/mjlab/asset_zoo/robots/i2rt_yam/xmls/assets/model2__15.stl": "ca8788184d54f30c18482cb9b48dffb010a93dbc029054b6cc515083a740d324",
"rl/mjlab/asset_zoo/robots/i2rt_yam/xmls/assets/model2__16.stl": "18bb99691eed524393e8f3a0ff29e79d97ab75689bc92144bbdfafa2cc79d37b",
"rl/mjlab/asset_zoo/robots/i2rt_yam/xmls/assets/model2__17.stl": "6513dbd192d2bf3a4f7016231f1ba5eca93916c9ebfdc334e9b435ca1d302ff6",
"rl/mjlab/asset_zoo/robots/i2rt_yam/xmls/assets/model2__2.stl": "30dab41f14bae95d737abca3c934eb5d570ab107dff0b40952fbdfb9d07eab79",
"rl/mjlab/asset_zoo/robots/i2rt_yam/xmls/assets/model2__3.stl": "9fd96f5ad387fe7dd96eaa65e386637800e2f5637b6167fbad4ecb37d8598f47",
"rl/mjlab/asset_zoo/robots/i2rt_yam/xmls/assets/model2__4.stl": "33c0e7ef538e2c4473efa9c4ef2a8441b6094b65090b4dec71abe82001d974df",
"rl/mjlab/asset_zoo/robots/i2rt_yam/xmls/assets/model2__5.stl": "09f08e7dead2d6363618471adadb1b94af39e8056676670f3fd21ec06c197b20",
"rl/mjlab/asset_zoo/robots/i2rt_yam/xmls/assets/model2__6.stl": "888b5919d10dbfb59b964c3cafc2ddbdac3b5426e20dd13bcaf4baedd5ee15c9",
"rl/mjlab/asset_zoo/robots/i2rt_yam/xmls/assets/model2__7.stl": "2d64b18f1e70b0460eb53335d340b7c879976a3093ec67b07bcb7245e371a91d",
"rl/mjlab/asset_zoo/robots/i2rt_yam/xmls/assets/model2__8.stl": "db3708da6b14daf3fc9a1909a8325b68f65ff69146cdd20b69a9ddc4ce51f50d",
"rl/mjlab/asset_zoo/robots/i2rt_yam/xmls/assets/model2__9.stl": "74969d09b8b7f75cc984c5c7e5526d2941c6724f48b3efa1fa8730cc7820f4f5",
"rl/mjlab/asset_zoo/robots/i2rt_yam/xmls/yam.xml": "79078fe009fd31c90343d5240b6f7df475ea30848aa994108b90d342e5b37d06",
"rl/mjlab/asset_zoo/robots/i2rt_yam/yam_constants.py": "3aa54282545d9c42534e9410ceedfda08e0f199addf14a64d7322aafa7843c2e",
"rl/mjlab/asset_zoo/robots/unitree_g1/__init__.py": "c11e649d6d5209becaf6dbca2d09dee83c17844b5d941c2e1cc6f16ba618bf14",
"rl/mjlab/asset_zoo/robots/unitree_g1/g1_constants.py": "30805e420c1dd03518ee582465fbce9cf097f4dc2eb0c1c2fce55c0332040c07",
"rl/mjlab/asset_zoo/robots/unitree_g1/xmls/assets/head_link.STL": "005fb67fbd3eff94aa8bf4a6e83238174e9f91b6721f7111594322f223724411",
"rl/mjlab/asset_zoo/robots/unitree_g1/xmls/assets/left_ankle_pitch_link.STL": "d49e3abc6f5b12e532062cd575b87b5ef40cd2a3fc18f54a1ca5bba4f773d51d",
"rl/mjlab/asset_zoo/robots/unitree_g1/xmls/assets/left_ankle_roll_link.STL": "c4092af943141d4d9f74232f3cfa345afc6565f46a077793b8ae0e68b39dc33f",
"rl/mjlab/asset_zoo/robots/unitree_g1/xmls/assets/left_elbow_link.STL": "fa752198accd104d5c4c3a01382e45165b944fbbc5acce085059223324e5bed3",
"rl/mjlab/asset_zoo/robots/unitree_g1/xmls/assets/left_hip_pitch_link.STL": "4725168105ee768ee31638ef22b53f6be2d7641bfd7cfefe803488d884776fa4",
"rl/mjlab/asset_zoo/robots/unitree_g1/xmls/assets/left_hip_roll_link.STL": "91f25922ee8a7c3152790051bebad17b4d9cd243569c38fe340285ff93a97acf",
"rl/mjlab/asset_zoo/robots/unitree_g1/xmls/assets/left_hip_yaw_link.STL": "a16d88aa6ddac8083aa7ad55ed317bea44b1fa003d314fba88965b7ed0f3b55b",
"rl/mjlab/asset_zoo/robots/unitree_g1/xmls/assets/left_knee_link.STL": "8d92b9e3d3a636761150bb8025e32514c4602b91c7028d523ee42b3e632de477",
"rl/mjlab/asset_zoo/robots/unitree_g1/xmls/assets/left_rubber_hand.STL": "cff2221a690fa69303f61fce68f2d155c1517b52efb6ca9262dd56e0bc6e70fe",
"rl/mjlab/asset_zoo/robots/unitree_g1/xmls/assets/left_shoulder_pitch_link.STL": "f0d1cfd02fcf0d42f95e678eeca33da3afbcc366ffba5c052847773ec4f31d52",
"rl/mjlab/asset_zoo/robots/unitree_g1/xmls/assets/left_shoulder_roll_link.STL": "fb9df21687773522598dc384f1a2945c7519f11cbc8bd372a49170316d6eee88",
"rl/mjlab/asset_zoo/robots/unitree_g1/xmls/assets/left_shoulder_yaw_link.STL": "1aa97e9748e924336567992181f78c7cd0652fd52a4afcca3df6b2ef6f9e712e",
"rl/mjlab/asset_zoo/robots/unitree_g1/xmls/assets/left_wrist_pitch_link.STL": "b251d8e05047f695d0f536cd78c11973cfa4e78d08cfe82759336cc3471de3a9",
"rl/mjlab/asset_zoo/robots/unitree_g1/xmls/assets/left_wrist_roll_link.STL": "edc387c9a0ba8c2237e9b296d32531426fabeb6f53e58df45c76106bca74148c",
"rl/mjlab/asset_zoo/robots/unitree_g1/xmls/assets/left_wrist_yaw_link.STL": "83f8fb3a726bf9613d65dd14f0f447cb918c3c95b3938042a0c9c09749267d3b",
"rl/mjlab/asset_zoo/robots/unitree_g1/xmls/assets/logo_link.STL": "8571a0a19bc4916fa55f91449f51d5fdefd751000054865a842449429d5f155b",
"rl/mjlab/asset_zoo/robots/unitree_g1/xmls/assets/pelvis.STL": "5ba6bbc888e630550140d3c26763f10206da8c8bd30ed886b8ede41c61f57a31",
"rl/mjlab/asset_zoo/robots/unitree_g1/xmls/assets/pelvis_contour_link.STL": "5cc5c2c7a312329e3feeb2b03d3fc09fc29705bd01864f6767e51be959662420",
"rl/mjlab/asset_zoo/robots/unitree_g1/xmls/assets/right_ankle_pitch_link.STL": "15be426539ec1be70246d4d82a168806db64a41301af8b35c197a33348c787a9",
"rl/mjlab/asset_zoo/robots/unitree_g1/xmls/assets/right_ankle_roll_link.STL": "4b66222ea56653e627711b56d0a8949b4920da5df091da0ceb343f54e884e3a5",
"rl/mjlab/asset_zoo/robots/unitree_g1/xmls/assets/right_elbow_link.STL": "1be925d7aa268bb8fddf5362b9173066890c7d32092c05638608126e59d1e2ab",
"rl/mjlab/asset_zoo/robots/unitree_g1/xmls/assets/right_hip_pitch_link.STL": "e4f3c99d7f4a7d34eadbef9461fc66e3486cb5442db1ec50c86317d459f1a9c6",
"rl/mjlab/asset_zoo/robots/unitree_g1/xmls/assets/right_hip_roll_link.STL": "4c254ef66a356f492947f360dd931965477b631e3fcc841f91ccc46d945d54f6",
"rl/mjlab/asset_zoo/robots/unitree_g1/xmls/assets/right_hip_yaw_link.STL": "e479c2936ca2057e9eb2f7dff6c189b7419d7b8484dea0b298cbb36a2a6aa668",
"rl/mjlab/asset_zoo/robots/unitree_g1/xmls/assets/right_knee_link.STL": "63c4008449c9bbe701a6e2b557b7a252e90cf3a5abcf54cee46862b9a69f8ec8",
"rl/mjlab/asset_zoo/robots/unitree_g1/xmls/assets/right_rubber_hand.STL": "99533b778bca6246144fa511bb9d4e555e075c641f2a0251e04372869cd99d67",
"rl/mjlab/asset_zoo/robots/unitree_g1/xmls/assets/right_shoulder_pitch_link.STL": "24cdb387e0128dfe602770a81c56cdce3a0181d34d039a11d1aaf8819b7b8c02",
"rl/mjlab/asset_zoo/robots/unitree_g1/xmls/assets/right_shoulder_roll_link.STL": "962b97c48f9ce9e8399f45dd9522e0865d19aa9fd299406b2d475a8fc4a53e81",
"rl/mjlab/asset_zoo/robots/unitree_g1/xmls/assets/right_shoulder_yaw_link.STL": "a0b76489271da0c72461a344c9ffb0f0c6e64f019ea5014c1624886c442a2fe5",
"rl/mjlab/asset_zoo/robots/unitree_g1/xmls/assets/right_wrist_pitch_link.STL": "d22f8f3b3127f15a63e5be1ee273cd5075786c3142f1c3d9f76cbf43d2a26477",
"rl/mjlab/asset_zoo/robots/unitree_g1/xmls/assets/right_wrist_roll_link.STL": "a7ee9212ff5b94d6cb7f52bb1bbf3f352194d5b598acff74f4c77d340c5b344f",
"rl/mjlab/asset_zoo/robots/unitree_g1/xmls/assets/right_wrist_yaw_link.STL": "bc9dece2d12509707e0057ba2e48df8f3d56db0c79410212963a25e8a50f61a6",
"rl/mjlab/asset_zoo/robots/unitree_g1/xmls/assets/torso_link_rev_1_0.STL": "11ddb46f2098efbbd8816b1d65893632d8e78be936376c7cdcd6771899ccc723",
"rl/mjlab/asset_zoo/robots/unitree_g1/xmls/assets/waist_roll_link_rev_1_0.STL": "b67c347a05abc3e8ddae600d98a082bee273bb39f8e651647708b0a7140a8a97",
"rl/mjlab/asset_zoo/robots/unitree_g1/xmls/assets/waist_yaw_link_rev_1_0.STL": "ec6db442b11f25eed898b5add07940c85d804f300de24dcbd264ccd8be7d554c",
"rl/mjlab/asset_zoo/robots/unitree_g1/xmls/g1.xml": "febdcbeffbbf84051556ae41a5ac1b43fb479a5d76bdb3f54824dbc2721c20aa",
"rl/mjlab/asset_zoo/robots/unitree_go1/__init__.py": "976aec5cf5633d29630ae1ef74d1139504a22cdffc2ef18ba1617ae679144f2f",
"rl/mjlab/asset_zoo/robots/unitree_go1/go1_constants.py": "fca9478fa292b4a0d7c473be31d6537ae613df7490f4941a885ad31d9723f565",
"rl/mjlab/asset_zoo/robots/unitree_go1/xmls/assets/calf.stl": "ad3685e132454fd64e4ca3e5089f0d812fb53958e95c14913a22e3193e9e9ed8",
"rl/mjlab/asset_zoo/robots/unitree_go1/xmls/assets/hip.stl": "b6599b9dd2751ebd87449203223f5683a6c235268b76fcc443e00e0d974a06c0",
"rl/mjlab/asset_zoo/robots/unitree_go1/xmls/assets/thigh.stl": "bbf7c0a441323088c59276dbec8a32e708a1c61f42d30a6f6775176366ea23f8",
"rl/mjlab/asset_zoo/robots/unitree_go1/xmls/assets/thigh_mirror.stl": "1a90b89de47a57cb12a2d703f6b008a6ace8a1890a381c27723296fd0d00b9d8",
"rl/mjlab/asset_zoo/robots/unitree_go1/xmls/assets/trunk.stl": "051a57995dc6f892fc9bc53cd94e9a894f3dc209fd646e94d55d764150013970",
"rl/mjlab/asset_zoo/robots/unitree_go1/xmls/go1.xml": "2f1856b0c83994c009cafbad1f5eac38d09efa1a64db73681ff72c9237822db3",
"rl/mjlab/entity/__init__.py": "2718cbce89ff2c6549d855d096329a72a87d5d77e92daa18b257ee946489cba6",
"rl/mjlab/entity/data.py": "03deabf1afc122192f031ef71a328e1443d602801d70da5cefe4b754980cfb71",
"rl/mjlab/entity/entity.py": "f0353cfc1e689d6e4cf0dc8d59d26847367cc90842ecebd9bc1a5a3044613eda",
"rl/mjlab/entity/variants.py": "f3c3ca5c2712f87a1647b3e9a27bf5d538a6e28a60f3e871ce2d8be65474e41d",
"rl/mjlab/envs/__init__.py": "5c41a9f8cfb59196dbd322e81a54df719d74fd30ff4e1e1958b5778ebc9831a1",
"rl/mjlab/envs/manager_based_rl_env.py": "e05034969baaa6ca38d7587190eef95d468fd413f182edd2c8d387c2950701f1",
"rl/mjlab/envs/mdp/__init__.py": "69d5a26435e69f8636093ef71a977bcb4f45ded4a4277c80c60defda41be1da9",
"rl/mjlab/envs/mdp/actions/__init__.py": "af023e77e5e624d86121742acab67a3eae2d67a7d2f1c595c9a59ba168386859",
"rl/mjlab/envs/mdp/actions/actions.py": "9cae83a2b0006320a8e15e040a0edbf5b4ae76d023832563f1ac36630f3b28d1",
"rl/mjlab/envs/mdp/actions/differential_ik.py": "b8b9654921afb7a46abb57719e2429c2274653296ea345469df7b92bc101124d",
"rl/mjlab/envs/mdp/curriculums.py": "844336b87cf71dfc6d6e60ba8001fdae42736deb9c8cd3e2cb81c1af0f7cb652",
"rl/mjlab/envs/mdp/dr/__init__.py": "633547b48443bffb3e0a04db289559e684e7c75b6c7becb069a82886930b04e1",
"rl/mjlab/envs/mdp/dr/_core.py": "166781592b42e07845acb9e836036b57d377c9b2ef8bc062a7059f6e3b5da41e",
"rl/mjlab/envs/mdp/dr/_types.py": "0dadb1cbf88bdf97cd92f636742715520cc5d27c7a6b0a2863a1be703c8c044b",
"rl/mjlab/envs/mdp/dr/actuator.py": "20ecdff1cb0d590726b1f5ed9d4e9d1de5a6380ca51ec7b05c0cf2ce34812b9f",
"rl/mjlab/envs/mdp/dr/body.py": "87a1135a03648880438270467301ff78b47297d1ec74755734b0859c12f6e15d",
"rl/mjlab/envs/mdp/dr/camera.py": "ee102f4a08fe984765cd6744c44b9a9a50f99229abc498002338b992835b38db",
"rl/mjlab/envs/mdp/dr/geom.py": "634dfe88eb5118d48708b9851cb42e9c9b38bf83b2f11b194686b644806f66a1",
"rl/mjlab/envs/mdp/dr/joint.py": "8ca0f9626b5c8b5fedb57013d3ed229ca00fa397378cfde329ebbfe1b2617e19",
"rl/mjlab/envs/mdp/dr/light.py": "26bf5a2209bd381319bff624b572918b5966e18c876851cccb0f7b0032710b24",
"rl/mjlab/envs/mdp/dr/material.py": "582ce46c775deb750648970c279fc0f7a3c0e8d021d99286fa27bb571b5c4b45",
"rl/mjlab/envs/mdp/dr/pair.py": "d1f99a5f200ff29da98761330a00a6e4cda9cc8073191697fc85b4ce0b46115d",
"rl/mjlab/envs/mdp/dr/site.py": "a3ed4a1d072a7417653d37f5ddcd8474278901b57f5309d0955d335985bec694",
"rl/mjlab/envs/mdp/dr/tendon.py": "eaf357196128be8c06af0e79715981509bd08ae427c367cdb0c6564c8f5ce42e",
"rl/mjlab/envs/mdp/events.py": "922a2475e58fe5ef15bb8a4dbae43608b874b9fb25cf12c0d794e6682ef646ad",
"rl/mjlab/envs/mdp/metrics.py": "5a0d5582aa7fe8c2d530dc5afdf151eace17d529429176a2fbff003d7d60e1ba",
"rl/mjlab/envs/mdp/observations.py": "610b2682fada5fbd7784416e53ba3375cd9e480542737ee4bd5b1f5af79b4b72",
"rl/mjlab/envs/mdp/rewards.py": "9a5908d02ee9ab05390961f70d021fb5762c03db53413b2bbe1fe8ce0699697a",
"rl/mjlab/envs/mdp/terminations.py": "3c0d64bdc9aa69d8fd372f343e47ab6f9da59a77d2efd8e3d860eff2c1ebf68a",
"rl/mjlab/envs/types.py": "9508db1d92998c2465bd24ff24b6966f5634a7b0ad7ecb0fcdb9df6e6fdf93d0",
"rl/mjlab/licenses/IsaacLab.txt": "df351ebfa8af317bf5a9ad83df63a8b68e4f5c04b71f30767cdb27aa434eac62",
"rl/mjlab/licenses/mujoco_menagerie-i2rt_yam.txt": "dc46c3a7dc48221aa783e634d15db0f816ba83ba5cdb77bbaa91525c61c7552f",
"rl/mjlab/licenses/mujoco_menagerie-unitree_g1.txt": "98335465f43a20b5850e4651db6e74c4aa1e9fc8e8813d38f345178045c0da50",
"rl/mjlab/licenses/mujoco_menagerie-unitree_go1.txt": "2b794acf1e250f5545ae5a47d81e5685ea5488ce077d1087ed8be5f81e9ba5c0",
"rl/mjlab/managers/__init__.py": "6fd6c581b1f283290e1e03d8a9feb5516a4e505763885e870ea47af38ddf20ae",
"rl/mjlab/managers/action_manager.py": "fec79f933fd9437dc9fcb784e4f214b9bae4aad2252c2d2d1387b11b605a5332",
"rl/mjlab/managers/command_manager.py": "e9cf2e6697722af77fa32060844819287e3a12fffb091234514b74c18762342a",
"rl/mjlab/managers/curriculum_manager.py": "cc34f5b0897ac148d1a51503a63d987e8b549549fcb1713cd35c12e20147cf0a",
"rl/mjlab/managers/event_manager.py": "c8d74e87aab629238ee6b4d0cb458e97a6938670fe492f76ce74bc0488401689",
"rl/mjlab/managers/manager_base.py": "90dedd602785e8a439a2eca5789137bb128674d181709a1be6ec42901a0326cd",
"rl/mjlab/managers/metrics_manager.py": "5c1f24e75fd21220b5c96f974127568150ce6e9ee16d0abb97dcd933173f6723",
"rl/mjlab/managers/observation_manager.py": "1be6b23b36b595e359f4b7d055dc51f59885579fc904767b5aa499f08e6d14fe",
"rl/mjlab/managers/recorder_manager.py": "9c30862b75b2c2a77b49f331125be382912099147c6ae982aa682a58ce7c3e66",
"rl/mjlab/managers/reward_manager.py": "c43705a88730a16b5dc9031367cfe423259fa1d752bf85aa20165d5e10174f04",
"rl/mjlab/managers/scene_entity_config.py": "12f0ac1baf8894a6435b2733f148cb0282dab561e481e657e7324a40c1e7fe92",
"rl/mjlab/managers/termination_manager.py": "87cc51aa9738e6d137ee936317d76d0b7472c24c58776b4f99dd1bde81d29a6a",
"rl/mjlab/py.typed": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
"rl/mjlab/rl/__init__.py": "d67a01c3df70b1d481498bf8f258c3c7f589e9d163605a7dc70fee83e9189597",
"rl/mjlab/rl/config.py": "7d391cc0d7a37a4f86c3c4f4a0adc6d75c546273085feed0ddf597aedba6134a",
"rl/mjlab/rl/exporter_utils.py": "4841e3c17518e51286670d6e5e7ff11ec0faec12a6555216312d442f9c1c33be",
"rl/mjlab/rl/runner.py": "61e9200bf304e3df71096cd69f28f4d5973157e61cdc549fc21b6f156326f77e",
"rl/mjlab/rl/spatial_softmax.py": "413922be88fe797c528f59409f9975979cd1886d8ed51b17e28b9377ceadec7e",
"rl/mjlab/rl/vecenv_wrapper.py": "d458aa421d72c979d0269a24e51a4b49a02e4ec8cb5a9d10e47290aa5d86a3e2",
"rl/mjlab/scene/__init__.py": "ff411f5da07d192bd797040453398eb68f5577c35e41b1e8f41aef58a02b9610",
"rl/mjlab/scene/scene.py": "42893e03dc3cc3ad2199831f5b70331bfe8be56cd0b785e07b8dad6b8fb73897",
"rl/mjlab/scene/scene.xml": "0fa9355112e117328986cda47e4fbf4f5d47904b7fccb805d70dccaccdd038a4",
"rl/mjlab/scripts/_cli.py": "bcb1f4394126ee112578c905fb4ba89d8f8e67c9e7df55205e0c425ab3759219",
"rl/mjlab/scripts/csv_to_npz.py": "ea412e75427b61150eb9a8bb72bdbc0ecb0c42b062538195cf360cd53cfc2d7a",
"rl/mjlab/scripts/demo.py": "b20dec8c1e9c71fec5929294d242ea34a9e2fc7f81fb384346b13a5dcfa6917a",
"rl/mjlab/scripts/export_scene.py": "da4519902a6e31a7883105d596521fd4db0bb1a14f55764c603e944a2ee2469b",
"rl/mjlab/scripts/gcs.py": "521bfdbd95156026ec568a11e3da24caab3826d9ece8852d3a8d015fc6d82143",
"rl/mjlab/scripts/list_envs.py": "0fd1214a8be3e0e757a620f4d547b6b733dee1203aac053c3e4bcf31bdc32e91",
"rl/mjlab/scripts/nan_viz.py": "2766d372e6ae5f2250b2b781e1307a3ec26f87d7884a1a28a56b016a10aa0826",
"rl/mjlab/scripts/play.py": "606f537c568247c5e9ba2e94ac1a197675411b360fb1db3231f1b46710539456",
"rl/mjlab/scripts/train.py": "25da18c990b2267dfa14db97b10397caa089fe9fafa071a614e166cbb75af0f7",
"rl/mjlab/scripts/visualize_terrain.py": "8e260e01a2abf5acb1ca73f26e9e9aef668132a4bb2b934730ea44b95317c588",
"rl/mjlab/sensor/__init__.py": "7961da9847ef0fe23e107bccf0b795cda90008e456e7dc85dfed000f86bfb332",
"rl/mjlab/sensor/builtin_sensor.py": "401e495124fa96a89c87cd510da8f43c693db3f9110b93000dcb515171e57adf",
"rl/mjlab/sensor/camera_sensor.py": "52e9761f838a224ea71c276aeddc11af47bee7050664fe0a88ede724471f1597",
"rl/mjlab/sensor/contact_sensor.py": "8b55ab0d6a21d0256e2fc29187c00cbfe4412cd08993050386c6a1f08197fac7",
"rl/mjlab/sensor/raycast_sensor.py": "9e714ac82a7916d1c9bc0f7c15fb209834a5c2b3c6198e4dc3f4c172127263d6",
"rl/mjlab/sensor/sensor.py": "c700983150dbd74fbceec7486b842af4dc8917a251af0277436aee09a69dd077",
"rl/mjlab/sensor/sensor_context.py": "9cb90de6639bcaf38756a584ea9c34f6b35626a5ccaa4a44a3db435f8a7600ca",
"rl/mjlab/sensor/terrain_height_sensor.py": "18a7b190a01fe57cf1e2dcf597dedc3da410eef1d571f4ae000edb6a68e5d692",
"rl/mjlab/sim/__init__.py": "6c7e8c6d5bd848c06b243d54522ca446a94e023235c0c8cb8dd53e13551cf3e2",
"rl/mjlab/sim/randomization.py": "96762d28662fea35c410245df8281fd45794ace4b69a2765f873d4c285dece94",
"rl/mjlab/sim/sim.py": "73ba9898c14cbac674e1e08cec0bdac8e22289c84d17c783e67daefcbc7709b8",
"rl/mjlab/sim/sim_data.py": "fe0191280f56e6a51c39f94363c1306d9dda65588a05a96177f6cf2ff8578667",
"rl/mjlab/tasks/__init__.py": "310ff3bbeed45f4792394e36ff98eb4b5b70e0657c2e5ef67dd363d510133d9b",
"rl/mjlab/tasks/cartpole/__init__.py": "161dd6cccab68333cd14fa206aba646b78ea2fb0ed191e59bdc943f9e761bd5a",
"rl/mjlab/tasks/cartpole/cartpole.xml": "c4adb6759cea1f380d8b20c067339976ca21a5d824a515a3cd5f75e0969bdbe1",
"rl/mjlab/tasks/cartpole/cartpole_env_cfg.py": "4523135d867e3ba3492147d056c3916f1db427d57866f57e9ac0a8f3bdaec7bc",
"rl/mjlab/tasks/manipulation/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
"rl/mjlab/tasks/manipulation/config/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
"rl/mjlab/tasks/manipulation/config/yam/__init__.py": "9f5978bf241f6890a2b64459c8b365818ab22068c1100b0f97d1de67a64af36c",
"rl/mjlab/tasks/manipulation/config/yam/env_cfgs.py": "d6358310e5a97e3d0254b249b6e19be9a1e30e42ed8ef348f3a0d6c11fdd7b5b",
"rl/mjlab/tasks/manipulation/config/yam/rl_cfg.py": "c013e8f318d51c3cacf6198986dcd5c3781d10a618165d9681b54e9975d25cbe",
"rl/mjlab/tasks/manipulation/lift_cube_env_cfg.py": "093b235694fb99d8c2bea79f9bf1ad0bc3078f0c97b981b46b8c10488206b390",
"rl/mjlab/tasks/manipulation/mdp/__init__.py": "02b2ea12a27078ba3ab5c444da0cdb461a765b0fec17f06d9120c4f767568802",
"rl/mjlab/tasks/manipulation/mdp/commands.py": "a8f9ccb2040f8e82452d4c82c1d7caedc41b4e1cbdac740da16202c2d0e26c26",
"rl/mjlab/tasks/manipulation/mdp/observations.py": "ef3a2dae26b4a626990474881d2daef485c9c7c26b78b0e873430767a2f04065",
"rl/mjlab/tasks/manipulation/mdp/rewards.py": "716b20339ffbdfdd839b28f855da257ff3cada2db587bfed995fe483f3674345",
"rl/mjlab/tasks/manipulation/mdp/terminations.py": "8da09eb99b1365c7e6bd5f7cf68e6b77a57da2c93975b265d6f211c757186186",
"rl/mjlab/tasks/manipulation/rl/__init__.py": "52e7cb371a86b6f62d3d0df8ad4fe6ac08875858b2e0ae11d33ac401935628f6",
"rl/mjlab/tasks/manipulation/rl/runner.py": "c9c43b3b8aebc444d655f40a5caa1ef145a8ce48745b7f08ce53eae1bd09ec37",
"rl/mjlab/tasks/registry.py": "f1c82c8f1ff5d24263787f4cfc2805b7b4103253fe59c054bf649d24c7cc9476",
"rl/mjlab/tasks/tracking/__init__.py": "269e19fac2c24adc6064d7c868741fa265f12d78ca07dd12ef7333febde21adf",
"rl/mjlab/tasks/tracking/config/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
"rl/mjlab/tasks/tracking/config/g1/__init__.py": "fe1f3b8b0fb27d8a353f7d5875b1624c3bdbe04802e2f65c56231b3ccb72e590",
"rl/mjlab/tasks/tracking/config/g1/env_cfgs.py": "ce67f0a0e4f8246aa13e7a1e855b30cbd5f5c40fec7ee8ee1c92ddf72ac44d50",
"rl/mjlab/tasks/tracking/config/g1/rl_cfg.py": "88a8a0acc8052a4bb64957a93ab2a88fa345ee84f63e405ffc6fa70148c6e104",
"rl/mjlab/tasks/tracking/mdp/__init__.py": "1c2637ed117786d35fc7a8b72fa36e11134268d8a2aaefa7109ed826c7eb75bf",
"rl/mjlab/tasks/tracking/mdp/commands.py": "b8700e65bb519b5e68962216ed78a107f6a3c02bab2ff75854e62b1983d1ad97",
"rl/mjlab/tasks/tracking/mdp/metrics.py": "d13a3af779454b911eea976e01115d11f35da64f14c8cf3a83c8a9b934f4c23a",
"rl/mjlab/tasks/tracking/mdp/observations.py": "af2cf6613d657529455eb254c7b13892ab67a0cf8ea5aec2d4df5fad6b4a5330",
"rl/mjlab/tasks/tracking/mdp/rewards.py": "fd7ef49a8015ca19c12586c6b87afca8c8ee7c03c3bb4b3f0a4b58c1dde6c7d4",
"rl/mjlab/tasks/tracking/mdp/terminations.py": "5f88576b5995408a744cde451c7e43982d2f03284cfeebb8856969ca3dc76019",
"rl/mjlab/tasks/tracking/rl/__init__.py": "95fe9fdc0cf223f6415d67607fbd4b2c6653df184cbb15f36403d0cb5ff7052b",
"rl/mjlab/tasks/tracking/rl/runner.py": "978b6f1c783d67bafd942d8790f1021108054441cdd6dd59be3f1c14262d705f",
"rl/mjlab/tasks/tracking/scripts/evaluate.py": "6ca116e3d5273f39acac3117e68871952c55a000c236e4dbcd762035cc99d742",
"rl/mjlab/tasks/tracking/tracking_env_cfg.py": "631b446ffc478c170ce5ef2a1a74b766e963f13a660f78b168060cd425cac243",
"rl/mjlab/tasks/velocity/__init__.py": "479a0ae48d56e95521d9cfd6ccb3bbcebc2bf985355d96879ce3a50d20a142ea",
"rl/mjlab/tasks/velocity/config/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
"rl/mjlab/tasks/velocity/config/booster_t1/__init__.py": "1a612197876f9187ace170cfd2e7c84f0e23232d5bfb578350185ca8e5e6eb40",
"rl/mjlab/tasks/velocity/config/booster_t1/env_cfgs.py": "eb2d8707af01380de8f718585b92912861f9e2172dc06fcfbaddd0e51baa1aa3",
"rl/mjlab/tasks/velocity/config/booster_t1/mdp.py": "14342b21816900df66c61223b9752f60ce383219abb411ce439ff24481e1d690",
"rl/mjlab/tasks/velocity/config/booster_t1/rl_cfg.py": "eafeaa715631bebb0cfbaf46e91fe299c66f62d3671726135911379010706b90",
"rl/mjlab/tasks/velocity/config/g1/__init__.py": "f4482cf1f1bb27989b0de3f1f45c0c1568185a632c9eb36cc18595cf95ac79cf",
"rl/mjlab/tasks/velocity/config/g1/env_cfgs.py": "5d891ef40987118d1bb2d4bb8791922387d8fdae22752f1e1420dc31f4ddeabf",
"rl/mjlab/tasks/velocity/config/g1/rl_cfg.py": "dcf4a23e031064085e081798c4d3b5bd1cda6ec58ad0081907f2e72c9142a131",
"rl/mjlab/tasks/velocity/config/go1/__init__.py": "9688ac3d4f925a59fcdfe5870753fb52ffc62c6944bb259bae6a3b7143f908cf",
"rl/mjlab/tasks/velocity/config/go1/env_cfgs.py": "3fe9a3f2882d0bb627c791fd009447911bf02dcb977d739ee87c26355909a7af",
"rl/mjlab/tasks/velocity/config/go1/rl_cfg.py": "ac9744e0a2c11db9e60f819c0e8423e6479e094c4aaebe189a6965b06bb90f63",
"rl/mjlab/tasks/velocity/mdp/__init__.py": "231f35d88929e89ae17dbf403b6d049dd9f1f8bba21bfaf74a087bb9435e7dfe",
"rl/mjlab/tasks/velocity/mdp/curriculums.py": "60a857b7e3456f48ba0ddcc33a70c154998a04ef215dbd40200cbdcdad4cea8b",
"rl/mjlab/tasks/velocity/mdp/observations.py": "4e56cefa6481c42e641902cc124cfdd02c3c92b45d8b294253aad0fc87ec9134",
"rl/mjlab/tasks/velocity/mdp/rewards.py": "da3a7b51d5d5c8c67493714e0a26ad29adc7c159e8df875f88bd4bcf5e1de4d4",
"rl/mjlab/tasks/velocity/mdp/terminations.py": "afebaf4170ed51e44e45626ed32e295d29dc43e96f93ca82ccb7b5c0c2c873cf",
"rl/mjlab/tasks/velocity/mdp/terrain_utils.py": "cb03bece16e1848746738404a4070454c4f753261b1c6537dfb1549c3ef7e627",
"rl/mjlab/tasks/velocity/mdp/velocity_command.py": "f20680098726b4f2811a02cbfe3391c946a493ef4536432762b63778e8a81231",
"rl/mjlab/tasks/velocity/rl/__init__.py": "998d67bb2c79979ed2999b14c74231e39a6d018f97c5429405b2991c36cdffa5",
"rl/mjlab/tasks/velocity/rl/runner.py": "05c396737a75720c09ea4495b6b500a5a21e66d9bf60162a2353245d54b5a91e",
"rl/mjlab/tasks/velocity/velocity_env_cfg.py": "8a272a8b3c81c4a0321027e404645c99a4143076522ed3521636bf293dd3137b",
"rl/mjlab/terrains/README.md": "2cf265575a9b76418306a1b4a3c053c3a4c04e181a6b37130b610efa22113021",
"rl/mjlab/terrains/__init__.py": "76c3c30dcc3027e9f8fc7929d9f473db22022eb2fc6cf68dcda556b74e30389a",
"rl/mjlab/terrains/config.py": "3005acad2ae9420be19c0c3c6a8954a190d70cbf19952e86e51c1ce7ba5d06ab",
"rl/mjlab/terrains/heightfield_terrains.py": "98ae7a098ac55a59c0958e3f040e1603395a87c643d93df89b7a44c1105eb08e",
"rl/mjlab/terrains/primitive_terrains.py": "8c526781a1e757fa7af1e165283d9b03f059298fc6704f4f147fdafaf60fd4b7",
"rl/mjlab/terrains/terrain_entity.py": "272a6e1c39f4221ded391f75792982c31f6ede8f44393f8a705d6acf83be9ca1",
"rl/mjlab/terrains/terrain_generator.py": "7f390739f8d074ce9b02b3e2df3d79780899b83096f9efa339d987aa533a4e65",
"rl/mjlab/terrains/utils.py": "a2a1650176f084b447d7c3a6a704dae17bc3a7c050f0cf4aa800c8430f64aa94",
"rl/mjlab/utils/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
"rl/mjlab/utils/actuator.py": "064435b65ac6f6c117c351882652ebe00a9cb36523d82f94a6aa9502276b1fbd",
"rl/mjlab/utils/buffers/__init__.py": "bfe46ad1bdc8c29555f2f6d675e703d91b4698d1cc735d90ffb86d23b120a929",
"rl/mjlab/utils/buffers/circular_buffer.py": "90713e0d41615b627337bfd245761e47f5650fa7373011cc6bfe0eaeb8b71809",
"rl/mjlab/utils/buffers/delay_buffer.py": "c9ab3fca0d384dce6d28478c8b0c1e48fc2a85b2a5180c4a392aa6937717b4d7",
"rl/mjlab/utils/color.py": "ac8c048df0c21ba48558f048d31a4298595105d02fefa1940b226d86f2188ef9",
"rl/mjlab/utils/gpu.py": "73995a8a2f38323463101e2db2621173682709d4347d610a94f62cdccf3c6e7a",
"rl/mjlab/utils/lab_api/__init__.py": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
"rl/mjlab/utils/lab_api/math.py": "85e267ee202dc8dc2b340be07ec7abe9b10da9b82c3bf33e16889f422fcf9ca8",
"rl/mjlab/utils/lab_api/string.py": "d0edaedbb1c0a6fed7dd4a72e1b0d8e7f2aa149bbc1ebc52547405e6f669371d",
"rl/mjlab/utils/lab_api/tasks/importer.py": "350f36e86224bcbec60bf427cb53c5fa1de111b34e3c489f3ab67ba766c7839d",
"rl/mjlab/utils/logging.py": "cbd7d5dceb22b84edc27a782925d080a0df2ee776df69d9df3d30c99af1a2200",
"rl/mjlab/utils/mujoco.py": "879b8109e03c4a32143899be4648ef02aa57125a40b36658d1fb361dd74216de",
"rl/mjlab/utils/nan_guard.py": "b3ba9d2c3f34c5e8f1c5249e28699141bd389971eed41cf8d05e0bf5f3199943",
"rl/mjlab/utils/noise/__init__.py": "cd43d801881ad37bc9d34c5263cf1f0c4aaad12355b1a565dfba5d2a722fd958",
"rl/mjlab/utils/noise/noise_cfg.py": "1132aa0bc88f498e5641504e9d79a26545513b123a526c91f61d5823693d54c8",
"rl/mjlab/utils/noise/noise_model.py": "1b6caa33ad612e512027072365ac2ecd3e761ae470cfff38ea945a7763da76bb",
"rl/mjlab/utils/os.py": "8364587683fd9c062e85a40302ed792b8ace389f64b6e0ebd5ec4dde1a9f6241",
"rl/mjlab/utils/random.py": "5ba445435efa5f064aac4b572d035c6a6e98ca3ce459faf1f00dc6f73cbabbc8",
"rl/mjlab/utils/spaces.py": "42eb28b6a48b06886df1c6b385243d340b14f89e71c0e407fda54e0907a53a60",
"rl/mjlab/utils/spec.py": "585b3fc90df15a12984b58e9fbc657950344a3cd45438fd73e35544c6d983719",
"rl/mjlab/utils/spec_config.py": "ac8cd1c7ec64a079d91a27b9f51c0dd4ded4cb7abf0fb85df1cca5acf5d6abcb",
"rl/mjlab/utils/string.py": "702c5ee5bc6b9a3fc5b27a1419daa0653301494bda9f7e5981eda898a601f8bc",
"rl/mjlab/utils/torch.py": "4ba058efa94cc779d2de54db969465338f365372b59943417bbce9193bf33536",
"rl/mjlab/utils/wandb.py": "f70e9108282c6fe115218ec0f5b6fcd0dc9f69b5a945ac2fe43f2e73fb47f155",
"rl/mjlab/utils/wrappers/__init__.py": "bd180f557c3fff4267fab0a7cc838c792a274b4f5d5e6590a3b694d2cad55fe6",
"rl/mjlab/utils/wrappers/video_recorder.py": "1e1dbf62726c711c0b4bd51c523ff0d89435a909a4a28f151c6a9ba959cfd29d",
"rl/mjlab/utils/xml.py": "5976ce00b501a39b42ac5605c2ed9bd05ea4feb5fbf1a2d778a1656ec0caaa3b",
"rl/mjlab/viewer/__init__.py": "6215875f50dd7fd1459682b707ce82da953f70dbcaedf0d363d55a49679b7d1f",
"rl/mjlab/viewer/base.py": "26da5b876828fd72387c351ef74ebe308c83beccae9b006d856780670a7dea29",
"rl/mjlab/viewer/debug_visualizer.py": "d96616c4d9f0ace6921350da19fb6d2d011675e67cba32d42510a6b4ba4d1b3a",
"rl/mjlab/viewer/model_sync.py": "bb23b1c23192a6c4f81364a45e8e0fc32c846d98fc1ec1b438fb1646748afd5c",
"rl/mjlab/viewer/native/__init__.py": "1cc4c390f38e36f2f708330fb39695a789d6ff5557503e5af64862ab2c29f928",
"rl/mjlab/viewer/native/keys.py": "30cf4d7872f7af74d9c8238289d22c6ffca82a56767debe54f4074ec41db9cf9",
"rl/mjlab/viewer/native/viewer.py": "95149f0266a546f9a9ba11215a6f13dcdafbe8324f38d07bbca49234d96a928c",
"rl/mjlab/viewer/native/visualizer.py": "4d91ee004f502740e726d44469073b1a2ab5dc5b94927ead0ab5a7cb9f6fd85a",
"rl/mjlab/viewer/offscreen_renderer.py": "4f8ecc4552413529b4559ea8112211f23603f78a091dd6491e75e324258660a6",
"rl/mjlab/viewer/viewer_config.py": "8040b01ebf951c59f56ec73f965cb9555b03dd11f711de49210c38109bb23154",
"rl/mjlab/viewer/viser/__init__.py": "ea76203023f96327501a320fe3012632d0bee61174152beb98fe3cb556ed71c0",
"rl/mjlab/viewer/viser/camera_viewer.py": "b60f0ed335e40f018847f172ce7e0f70c8ba152d61e2f25eda7092ac1e7fbca4",
"rl/mjlab/viewer/viser/overlays.py": "f6184f1571964fbc94c81a9a607548b4a7d691e54c72ca74e06cf978f9a2eb56",
"rl/mjlab/viewer/viser/reward_bar_panel.py": "bff4194a48f6deaf929ee4945d670d73946bb302e13bddca782179f6ec6a4872",
"rl/mjlab/viewer/viser/scene.py": "107bbbf30da399289161c9815ed6dd53d4e2aa889f56eb32dde822f734bea0b2",
"rl/mjlab/viewer/viser/term_plotter.py": "4b2694fb5633fae6c01e71cf5e2c05a68ad2e5bbe6579f9339296572b6c5371d",
"rl/mjlab/viewer/viser/viewer.py": "8f11f8a13f6240178dcd5ea00df7ca387fe9e136cef839393d8ebfd93138b17e",
"rl/rsl_rl/LICENSE": "3f58bb6b9272e416b8ccbe65889463f75a2fdf66e0e5722513f8622fb23dc778",
"rl/rsl_rl/__init__.py": "9b4c420fef3ae39683544833b6d2efbb26aecfc0583ec59be7f4eeb748fecf66",
"rl/rsl_rl/algorithms/__init__.py": "e370271e689bada5cdf2b4623e0d16d75e8439864a2ae913b22215f7b2985843",
"rl/rsl_rl/algorithms/distillation.py": "3d9fb8e010346ab42f3f4a84584e7e175a7632b39700e7778d0a01746a6317a7",
"rl/rsl_rl/algorithms/ppo.py": "01a82cea45fa9d8ccb161bd75e07d8623d25450ff92982e9c9d59ca8a9b78393",
"rl/rsl_rl/env/__init__.py": "33bfc9562580377c2ce6388c37b0df2b8e992c22ce46923443681cbbf1960d60",
"rl/rsl_rl/env/vec_env.py": "4a72df2fbf9b3e162e0e4b10e0b473511f02ad8c3ebaab0b7993ddfd63a9fa43",
"rl/rsl_rl/extensions/__init__.py": "20dc79170fb9777b7267abe4296b5c22f4b417d5a18c4346fb8159f9d4aea62f",
"rl/rsl_rl/extensions/rnd.py": "7ff5a34f54bcfe8e9494cf5c27155102b6f50c9b10dff49f7a1d98f32e6e16ca",
"rl/rsl_rl/extensions/symmetry.py": "4121d8804c9555118f031f604bc5232c85db0b41ce8d8295a5a22b85190b4048",
"rl/rsl_rl/licenses/LICENSE": "3f58bb6b9272e416b8ccbe65889463f75a2fdf66e0e5722513f8622fb23dc778",
"rl/rsl_rl/licenses/licenses/dependencies/codespell-license.txt": "8177f97513213526df2cf6184d8ff986c675afb514d4e68a404010521b880643",
"rl/rsl_rl/licenses/licenses/dependencies/furo-license.txt": "f68043bb5f4701df244347c82cd2fe24741a56c56a767a1d02bea053670a1ec4",
"rl/rsl_rl/licenses/licenses/dependencies/gitpython-license.txt": "b63a4621e0151014fea8bf668b04c9fd633ea16745d5041a29594ebb12a97ba4",
"rl/rsl_rl/licenses/licenses/dependencies/lucas-c-pre-commit-hooks-license.txt": "6465878993b3f6bbec852eceb213d8163731a637f3e422707377aadf557590f2",
"rl/rsl_rl/licenses/licenses/dependencies/neptune-license.txt": "c71d239df91726fc519c6eb72d318ec65820627232b2f796219e87dcf35d0ab4",
"rl/rsl_rl/licenses/licenses/dependencies/numpy-license.txt": "c17011467b4fe6e20fbfbeda4df09e58c2bca5480a063e55735a4000a379aad9",
"rl/rsl_rl/licenses/licenses/dependencies/onnx-license.txt": "cfc7749b96f63bd31c3c42b5c471bf756814053e847c10f3eb003417bc523d30",
"rl/rsl_rl/licenses/licenses/dependencies/onnxscript-license.txt": "2f07c72751aed99790b8a4869cf2311df85a860b22ded05fa22803587a48922c",
"rl/rsl_rl/licenses/licenses/dependencies/pre-commit-hooks-license.txt": "ea2ca27cba7cc35822d95a46d59bcd3cc88e196592e6390d1949a359ffc990e8",
"rl/rsl_rl/licenses/licenses/dependencies/pre-commit-license.txt": "ea2ca27cba7cc35822d95a46d59bcd3cc88e196592e6390d1949a359ffc990e8",
"rl/rsl_rl/licenses/licenses/dependencies/pytest-license.txt": "ca836a5f9ecca3b2f350230faa20a48fb8b145653b5568d784862df864706b9b",
"rl/rsl_rl/licenses/licenses/dependencies/ruff-license.txt": "aaa18222a10861f3210e4c843794bf87cc56ce62f6732caa52b9525c2117ab25",
"rl/rsl_rl/licenses/licenses/dependencies/setuptools-license.txt": "dc0589ce4ab317782b1bfe5b24d1a3528eed3c028be8f197c516997c777cdd0c",
"rl/rsl_rl/licenses/licenses/dependencies/sphinx-license.txt": "98eac23a35c127d59613f757846b6af90cd4137bbdde8bc3feaf38e28e92f913",
"rl/rsl_rl/licenses/licenses/dependencies/tensorboard-license.txt": "edba09b5279dd73272b22c60479221736076dc3866352712db760cc8179061a4",
"rl/rsl_rl/licenses/licenses/dependencies/tensordict-license.txt": "c5d8d2e3fc64f88c272ee2050af4d84e5f58f1a5d17a3aa99a47b77466a6fe72",
"rl/rsl_rl/licenses/licenses/dependencies/torch-license.txt": "6f598239866f2794b1170c870217a604b6acc98d4aeb4694d1a2f855da1f289c",
"rl/rsl_rl/licenses/licenses/dependencies/wandb-license.txt": "8b338a449a463b1d4fac988638a4741ec35d941e49747d9dd19e0fedaeecb137",
"rl/rsl_rl/licenses/licenses/dependencies/wheel-license.txt": "30c23618679108f3e8ea1d2a658c7ca417bdfc891c98ef1a89fa4ff0c9828654",
"rl/rsl_rl/licenses/pfrl.txt": "1d68cbdc99c9967f63a588ff73fb59461882b21508fe2f91a22f5cf3a4d7ef65",
"rl/rsl_rl/models/__init__.py": "6444fc1df8bc94d40ab1911f0e8e6b3f777459986e7f7944d25b7e5ca9b3e206",
"rl/rsl_rl/models/cnn_model.py": "bc8119274d2058f0abe3482563e6e391aba312fa19c59d961413cfe09f4b0d93",
"rl/rsl_rl/models/mlp_model.py": "86a8ea5c77fdae8167b84d817e5280fe226bf15d2aafab8455c6366155ab2eee",
"rl/rsl_rl/models/rnn_model.py": "80c1d6452ef021e4d7b174231ed7abac1ed0db8de9604561f8060148d4333ccc",
"rl/rsl_rl/modules/__init__.py": "336085b45fb1e4a807f29cb2703b63f933cdf2c2ff00279cef3cfed14bdccc49",
"rl/rsl_rl/modules/cnn.py": "4f5bab5f192c42180125f32a4af350108dd1c247e1e84c81f6cf4ea560037ccf",
"rl/rsl_rl/modules/distribution.py": "ef490fac4baa8a6e6c2b6541cddd4c17f8838097c4298fc7c7efdfefa119941f",
"rl/rsl_rl/modules/mlp.py": "bad934b364b26cd47b6d1612e00ace107a425ae4d272c0cd0c82cbfc57bbcdbc",
"rl/rsl_rl/modules/normalization.py": "67a70c88b540f212759c6e6c023aed41321ea9ad949e5faf428670fafbdf73bd",
"rl/rsl_rl/modules/rnn.py": "07f0eec06c7f2638edbdbc3ba25f56122a7941cb0543138f4a6b04675b2a97e6",
"rl/rsl_rl/runners/__init__.py": "26f4130a403414e13041b62a848d04da3fb0166757603950ea758b151d4b5aaa",
"rl/rsl_rl/runners/distillation_runner.py": "7445cbaf3e1fa241d742447ab17d941a659c92b6f8365ef22c5d818923d9e275",
"rl/rsl_rl/runners/on_policy_runner.py": "c2dbe47974b047e6992e37ff39ac0a236c8fa5cb595cb9e1715e1344e81dc852",
"rl/rsl_rl/storage/__init__.py": "7c9d96df3cad10aa27f34e77f7105a29f74e2d7dd10467b2919f99867a69298e",
"rl/rsl_rl/storage/rollout_storage.py": "d2728ab2cbc280c163a3d33289216d807ac60e1edf41fb85dd09560e966c9b04",
"rl/rsl_rl/utils/__init__.py": "b04e7ce7ccbbaefa275930e5ffc33edd84e77364a14d7e5717e6a5cb02338066",
"rl/rsl_rl/utils/log_writer.py": "53e9447d4337ab7e2d9f48f6ee8ee07ab865789ff1c722e2bc8b3f44ee247693",
"rl/rsl_rl/utils/logger.py": "ff9a592899c6eb803d4a434cef8712aa0424575ac01b13575f3efe2596d4116e",
"rl/rsl_rl/utils/neptune_log_writer.py": "971784c3291d7d7dc8e93202fb8d0102fa67214227440569719cbad278c2cf77",
"rl/rsl_rl/utils/utils.py": "37e1cf2961d91204633e3d3e46951c3abd695c9c4b48a269cbc97846ac155751",
"rl/rsl_rl/utils/wandb_log_writer.py": "69effd6926c8eb7ca4de96d3311e33ccef9cfe8556ca13fec9421e9c8f17ad8d",
"rl/upstream/import.json": "4d2a13b00f229418012b5acd9afa26f3d4fb7f0ab611905dce111392f19671a0",
"rl/upstream/mjlab-local.patch": "178cb15db535a057fd1a064b58f4658f9b0a35e119b1c08d7e37dbb833e4938c",
"rl/upstream/mjlab-pyproject.toml": "2ef132b218db99f26aef59e9df221b520003a83bd7fc4ad0112d9cada2ca9b44",
"rl/upstream/rsl_rl-METADATA": "de4a41da7206f3310ea8f1c9149a0a17859e72cdf4ec8e53888e7a89ef7bc9f3",
"rl/upstream/tests/test_booster_t1.py": "6c8c2549db932150fb87edf58ff8240fd539f4f90fdb9fb074ae5c8c9349a088",
"rl/upstream/tests/test_sim.py": "d6caca50e5e5d86556cd713f3497cd3dcb39a32556738883ea1e5fd08c8ab737",
"rl/upstream/tests/test_velocity_command.py": "4ed39d6345958a48d76db11562f64f4d84d5d2099ed391255f9d0dd78173ff99"
},
"schema_version": 1,
"upstream": {
"booster_assets": {
"commit": "12a97516f57469078707afb587a5f84c1cb110d1",
"model": "T1_23dof.xml",
"upstream": "https://github.com/BoosterRobotics/booster_assets.git"
},
"mjlab": {
"added_modules": [
"asset_zoo/robots/booster_t1",
"tasks/velocity/config/booster_t1"
],
"commit": "8bb2bc9e187797689ec82c78c35019afd2fd0376",
"inherited_patch": "rl/upstream/mjlab-local.patch",
"upstream": "https://github.com/mujocolab/mjlab",
"version": "1.6.0"
},
"rsl_rl": {
"installed_RECORD_verified": true,
"upstream": "https://github.com/leggedrobotics/rsl_rl",
"version": "5.5.1"
}
}
}
+202
View File
@@ -0,0 +1,202 @@
Apache License
Version 2.0, January 2004
http://www.apache.org/licenses/
TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION
1. Definitions.
"License" shall mean the terms and conditions for use, reproduction,
and distribution as defined by Sections 1 through 9 of this document.
"Licensor" shall mean the copyright owner or entity authorized by
the copyright owner that is granting the License.
"Legal Entity" shall mean the union of the acting entity and all
other entities that control, are controlled by, or are under common
control with that entity. For the purposes of this definition,
"control" means (i) the power, direct or indirect, to cause the
direction or management of such entity, whether by contract or
otherwise, or (ii) ownership of fifty percent (50%) or more of the
outstanding shares, or (iii) beneficial ownership of such entity.
"You" (or "Your") shall mean an individual or Legal Entity
exercising permissions granted by this License.
"Source" form shall mean the preferred form for making modifications,
including but not limited to software source code, documentation
source, and configuration files.
"Object" form shall mean any form resulting from mechanical
transformation or translation of a Source form, including but
not limited to compiled object code, generated documentation,
and conversions to other media types.
"Work" shall mean the work of authorship, whether in Source or
Object form, made available under the License, as indicated by a
copyright notice that is included in or attached to the work
(an example is provided in the Appendix below).
"Derivative Works" shall mean any work, whether in Source or Object
form, that is based on (or derived from) the Work and for which the
editorial revisions, annotations, elaborations, or other modifications
represent, as a whole, an original work of authorship. For the purposes
of this License, Derivative Works shall not include works that remain
separable from, or merely link (or bind by name) to the interfaces of,
the Work and Derivative Works thereof.
"Contribution" shall mean any work of authorship, including
the original version of the Work and any modifications or additions
to that Work or Derivative Works thereof, that is intentionally
submitted to Licensor for inclusion in the Work by the copyright owner
or by an individual or Legal Entity authorized to submit on behalf of
the copyright owner. For the purposes of this definition, "submitted"
means any form of electronic, verbal, or written communication sent
to the Licensor or its representatives, including but not limited to
communication on electronic mailing lists, source code control systems,
and issue tracking systems that are managed by, or on behalf of, the
Licensor for the purpose of discussing and improving the Work, but
excluding communication that is conspicuously marked or otherwise
designated in writing by the copyright owner as "Not a Contribution."
"Contributor" shall mean Licensor and any individual or Legal Entity
on behalf of whom a Contribution has been received by Licensor and
subsequently incorporated within the Work.
2. Grant of Copyright License. Subject to the terms and conditions of
this License, each Contributor hereby grants to You a perpetual,
worldwide, non-exclusive, no-charge, royalty-free, irrevocable
copyright license to reproduce, prepare Derivative Works of,
publicly display, publicly perform, sublicense, and distribute the
Work and such Derivative Works in Source or Object form.
3. Grant of Patent License. Subject to the terms and conditions of
this License, each Contributor hereby grants to You a perpetual,
worldwide, non-exclusive, no-charge, royalty-free, irrevocable
(except as stated in this section) patent license to make, have made,
use, offer to sell, sell, import, and otherwise transfer the Work,
where such license applies only to those patent claims licensable
by such Contributor that are necessarily infringed by their
Contribution(s) alone or by combination of their Contribution(s)
with the Work to which such Contribution(s) was submitted. If You
institute patent litigation against any entity (including a
cross-claim or counterclaim in a lawsuit) alleging that the Work
or a Contribution incorporated within the Work constitutes direct
or contributory patent infringement, then any patent licenses
granted to You under this License for that Work shall terminate
as of the date such litigation is filed.
4. Redistribution. You may reproduce and distribute copies of the
Work or Derivative Works thereof in any medium, with or without
modifications, and in Source or Object form, provided that You
meet the following conditions:
(a) You must give any other recipients of the Work or
Derivative Works a copy of this License; and
(b) You must cause any modified files to carry prominent notices
stating that You changed the files; and
(c) You must retain, in the Source form of any Derivative Works
that You distribute, all copyright, patent, trademark, and
attribution notices from the Source form of the Work,
excluding those notices that do not pertain to any part of
the Derivative Works; and
(d) If the Work includes a "NOTICE" text file as part of its
distribution, then any Derivative Works that You distribute must
include a readable copy of the attribution notices contained
within such NOTICE file, excluding those notices that do not
pertain to any part of the Derivative Works, in at least one
of the following places: within a NOTICE text file distributed
as part of the Derivative Works; within the Source form or
documentation, if provided along with the Derivative Works; or,
within a display generated by the Derivative Works, if and
wherever such third-party notices normally appear. The contents
of the NOTICE file are for informational purposes only and
do not modify the License. You may add Your own attribution
notices within Derivative Works that You distribute, alongside
or as an addendum to the NOTICE text from the Work, provided
that such additional attribution notices cannot be construed
as modifying the License.
You may add Your own copyright statement to Your modifications and
may provide additional or different license terms and conditions
for use, reproduction, or distribution of Your modifications, or
for any such Derivative Works as a whole, provided Your use,
reproduction, and distribution of the Work otherwise complies with
the conditions stated in this License.
5. Submission of Contributions. Unless You explicitly state otherwise,
any Contribution intentionally submitted for inclusion in the Work
by You to the Licensor shall be under the terms and conditions of
this License, without any additional terms or conditions.
Notwithstanding the above, nothing herein shall supersede or modify
the terms of any separate license agreement you may have executed
with Licensor regarding such Contributions.
6. Trademarks. This License does not grant permission to use the trade
names, trademarks, service marks, or product names of the Licensor,
except as required for reasonable and customary use in describing the
origin of the Work and reproducing the content of the NOTICE file.
7. Disclaimer of Warranty. Unless required by applicable law or
agreed to in writing, Licensor provides the Work (and each
Contributor provides its Contributions) on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
implied, including, without limitation, any warranties or conditions
of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A
PARTICULAR PURPOSE. You are solely responsible for determining the
appropriateness of using or redistributing the Work and assume any
risks associated with Your exercise of permissions under this License.
8. Limitation of Liability. In no event and under no legal theory,
whether in tort (including negligence), contract, or otherwise,
unless required by applicable law (such as deliberate and grossly
negligent acts) or agreed to in writing, shall any Contributor be
liable to You for damages, including any direct, indirect, special,
incidental, or consequential damages of any character arising as a
result of this License or out of the use or inability to use the
Work (including but not limited to damages for loss of goodwill,
work stoppage, computer failure or malfunction, or any and all
other commercial damages or losses), even if such Contributor
has been advised of the possibility of such damages.
9. Accepting Warranty or Additional Liability. While redistributing
the Work or Derivative Works thereof, You may choose to offer,
and charge a fee for, acceptance of support, warranty, indemnity,
or other liability obligations and/or rights consistent with this
License. However, in accepting such obligations, You may act only
on Your own behalf and on Your sole responsibility, not on behalf
of any other Contributor, and only if You agree to indemnify,
defend, and hold each Contributor harmless for any liability
incurred by, or claims asserted against, such Contributor by reason
of your accepting any such warranty or additional liability.
END OF TERMS AND CONDITIONS
APPENDIX: How to apply the Apache License to your work.
To apply the Apache License to your work, attach the following
boilerplate notice, with the fields enclosed by brackets "[]"
replaced with your own identifying information. (Don't include
the brackets!) The text should be enclosed in the appropriate
comment syntax for the file format. We also recommend that a
file or class name and description of purpose be included on the
same "printed page" as the copyright notice for easier
identification within third-party archives.
Copyright 2025, The mjlab Developers
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
+75
View File
@@ -0,0 +1,75 @@
import os
import sys
# Default to EGL for GPU-accelerated offscreen rendering on Linux. Must be set
# before any mujoco import: mujoco's gl_context module captures MUJOCO_GL once
# at load time. Override with e.g. MUJOCO_GL=osmesa on clusters without EGL.
# Linux-only because mujoco's gl_context rejects "egl" on macOS/Windows and
# raises at import. On those platforms we leave MUJOCO_GL alone so mujoco
# defaults to GLFW.
if sys.platform.startswith("linux"):
os.environ.setdefault("MUJOCO_GL", "egl")
import traceback
from importlib.metadata import entry_points
from pathlib import Path
import tyro
import warp as wp
MJLAB_SRC_PATH: Path = Path(__file__).parent
TYRO_FLAGS = (
# Don't let users switch between types in unions. This produces a simpler CLI
# with flatter helptext, at the cost of some flexibility. Type changes can
# just be done in code.
tyro.conf.AvoidSubcommands,
# Disable automatic flag conversion (e.g., use `--flag False` instead of
# `--no-flag` for booleans).
tyro.conf.FlagConversionOff,
# Use Python syntax for collections: --tuple (1,2,3) instead of --tuple 1 2 3.
# Helps with wandb sweep compatibility: https://brentyi.github.io/tyro/wandb_sweeps/
tyro.conf.UsePythonSyntaxForLiteralCollections,
)
def _configure_warp() -> None:
"""Configure Warp globally for mjlab."""
wp.config.enable_backward = False
# Keep warp verbose by default to show kernel compilation progress.
# Override with MJLAB_WARP_QUIET=1 environment variable if needed.
quiet = os.environ.get("MJLAB_WARP_QUIET", "0").lower() in ("1", "true", "yes")
wp.config.log_level = wp.LOG_WARNING if quiet else wp.LOG_INFO
def _import_registered_packages() -> None:
"""Auto-discover and import packages registered via entry points.
Looks for packages registered under the 'mjlab.tasks' entry point group.
Each discovered package is imported, which allows it to register custom
environments with gymnasium.
"""
mjlab_tasks = entry_points().select(group="mjlab.tasks")
for entry_point in mjlab_tasks:
try:
entry_point.load()
except Exception:
print(
f"[WARN] Failed to load task package '{entry_point.name}' ({entry_point.value}):",
file=sys.stderr,
)
traceback.print_exc(file=sys.stderr)
def _configure_mediapy() -> None:
"""Point mediapy at the bundled imageio-ffmpeg binary."""
import imageio_ffmpeg
import mediapy
mediapy.set_ffmpeg(imageio_ffmpeg.get_ffmpeg_exe())
_configure_warp()
_configure_mediapy()
_import_registered_packages()
+63
View File
@@ -0,0 +1,63 @@
"""Actuator implementations for mjlab."""
from mjlab.actuator.actuator import Actuator as Actuator
from mjlab.actuator.actuator import ActuatorCfg as ActuatorCfg
from mjlab.actuator.actuator import ActuatorCmd as ActuatorCmd
from mjlab.actuator.actuator import CommandField as CommandField
from mjlab.actuator.builtin_actuator import (
BuiltinDcMotorActuator as BuiltinDcMotorActuator,
)
from mjlab.actuator.builtin_actuator import (
BuiltinDcMotorActuatorCfg as BuiltinDcMotorActuatorCfg,
)
from mjlab.actuator.builtin_actuator import (
BuiltinMotorActuator as BuiltinMotorActuator,
)
from mjlab.actuator.builtin_actuator import (
BuiltinMotorActuatorCfg as BuiltinMotorActuatorCfg,
)
from mjlab.actuator.builtin_actuator import (
BuiltinMuscleActuator as BuiltinMuscleActuator,
)
from mjlab.actuator.builtin_actuator import (
BuiltinMuscleActuatorCfg as BuiltinMuscleActuatorCfg,
)
from mjlab.actuator.builtin_actuator import (
BuiltinPdActuator as BuiltinPdActuator,
)
from mjlab.actuator.builtin_actuator import (
BuiltinPdActuatorCfg as BuiltinPdActuatorCfg,
)
from mjlab.actuator.builtin_actuator import (
BuiltinPositionActuator as BuiltinPositionActuator,
)
from mjlab.actuator.builtin_actuator import (
BuiltinPositionActuatorCfg as BuiltinPositionActuatorCfg,
)
from mjlab.actuator.builtin_actuator import (
BuiltinVelocityActuator as BuiltinVelocityActuator,
)
from mjlab.actuator.builtin_actuator import (
BuiltinVelocityActuatorCfg as BuiltinVelocityActuatorCfg,
)
from mjlab.actuator.builtin_actuator import (
DcMotorDatasheetParams as DcMotorDatasheetParams,
)
from mjlab.actuator.builtin_actuator import (
DcMotorInputMode as DcMotorInputMode,
)
from mjlab.actuator.builtin_actuator import (
DcMotorPhysicalParams as DcMotorPhysicalParams,
)
from mjlab.actuator.builtin_group import BuiltinActuatorGroup as BuiltinActuatorGroup
from mjlab.actuator.dc_actuator import DcMotorActuator as DcMotorActuator
from mjlab.actuator.dc_actuator import DcMotorActuatorCfg as DcMotorActuatorCfg
from mjlab.actuator.fused_group import FusedActuatorGroup as FusedActuatorGroup
from mjlab.actuator.learned_actuator import LearnedMlpActuator as LearnedMlpActuator
from mjlab.actuator.learned_actuator import (
LearnedMlpActuatorCfg as LearnedMlpActuatorCfg,
)
from mjlab.actuator.pd_actuator import IdealPdActuator as IdealPdActuator
from mjlab.actuator.pd_actuator import IdealPdActuatorCfg as IdealPdActuatorCfg
from mjlab.actuator.xml_actuator import XmlActuator as XmlActuator
from mjlab.actuator.xml_actuator import XmlActuatorCfg as XmlActuatorCfg
+436
View File
@@ -0,0 +1,436 @@
"""Base actuator interface."""
from __future__ import annotations
import dataclasses
from abc import ABC, abstractmethod
from dataclasses import dataclass
from enum import Enum
from typing import TYPE_CHECKING, ClassVar, Generic, Literal, TypeVar
import mujoco
import mujoco_warp as mjwarp
import torch
from mjlab.utils.buffers import DelayBuffer
if TYPE_CHECKING:
from mjlab.entity import Entity
from mjlab.entity.data import EntityData
ActuatorCfgT = TypeVar("ActuatorCfgT", bound="ActuatorCfg")
CommandField = Literal["position", "velocity", "effort"]
class TransmissionType(str, Enum):
"""Transmission types for actuators."""
JOINT = "joint"
TENDON = "tendon"
SITE = "site"
@dataclass(kw_only=True)
class ActuatorCfg(ABC):
target_names_expr: tuple[str, ...]
"""Targets that are part of this actuator group.
Can be a tuple of names or tuple of regex expressions. Interpreted based on
transmission_type.
"""
transmission_type: TransmissionType = TransmissionType.JOINT
"""Transmission type. Defaults to JOINT."""
armature: float | None = None
"""Reflected rotor inertia. None preserves the XML value."""
frictionloss: float | None = None
"""Friction loss force limit. None preserves the XML value.
Applies a constant friction force opposing motion, independent of load or velocity.
Also known as dry friction or load-independent friction.
"""
viscous_damping: float | None = None
"""Passive viscous damping coefficient. None preserves the XML value.
Produces a dissipative force f(v) = -b·v proportional to velocity. Always present
regardless of actuator activity. Unlike ``damping`` (the PD derivative gain kv, which
is active control), this is a passive property.
Maps to ``<joint damping>`` for JOINT transmission and ``<tendon damping>``
for TENDON transmission. Ignored for SITE.
"""
delay_min_lag: int = 0
"""Minimum command delay in physics timesteps.
Each step, a lag is sampled uniformly from [min, max]. The command target arrives
that many steps late at the actuator's control law. Models communication and bus
latency between the policy and the motor (as opposed to observation delay, which
models sensor pipeline latency).
"""
delay_max_lag: int = 0
"""Maximum command delay in physics timesteps. Set > 0 to enable delay."""
delay_hold_prob: float = 0.0
"""Probability of keeping the current lag instead of resampling."""
delay_update_period: int = 0
"""How often to resample the lag, in physics timesteps (0 = every step)."""
delay_per_env_phase: bool = True
"""Stagger lag resampling across environments so they don't all update
on the same step."""
def __post_init__(self) -> None:
if self.armature is not None:
assert self.armature >= 0.0, "armature must be non-negative."
if self.frictionloss is not None:
assert self.frictionloss >= 0.0, "frictionloss must be non-negative."
if self.viscous_damping is not None:
assert self.viscous_damping >= 0.0, "viscous_damping must be non-negative."
if self.transmission_type == TransmissionType.SITE:
if (
(self.armature is not None and self.armature > 0.0)
or (self.frictionloss is not None and self.frictionloss > 0.0)
or (self.viscous_damping is not None and self.viscous_damping > 0.0)
):
raise ValueError(
f"{self.__class__.__name__}: armature, frictionloss, and viscous_damping are "
"not supported for SITE transmission type."
)
assert self.delay_min_lag >= 0, "delay_min_lag must be non-negative."
assert self.delay_max_lag >= 0, "delay_max_lag must be non-negative."
assert self.delay_min_lag <= self.delay_max_lag, (
"delay_min_lag must be <= delay_max_lag."
)
assert 0.0 <= self.delay_hold_prob <= 1.0, "delay_hold_prob must be in [0, 1]."
assert self.delay_update_period >= 0, "delay_update_period must be non-negative."
@abstractmethod
def build(
self, entity: Entity, target_ids: list[int], target_names: list[str]
) -> Actuator:
"""Build actuator instance.
Args:
entity: Entity this actuator belongs to.
target_ids: Local target indices (for indexing entity arrays).
target_names: Target names corresponding to target_ids.
Returns:
Actuator instance.
"""
raise NotImplementedError
@dataclass
class ActuatorCmd:
"""High-level actuator command with targets and current state.
Passed to actuator's `compute()` method to generate low-level control signals.
All tensors have shape (num_envs, num_targets).
"""
position_target: torch.Tensor
"""Desired positions (joint positions, tendon lengths, or site positions)."""
velocity_target: torch.Tensor
"""Desired velocities (joint velocities, tendon velocities, or site velocities)."""
effort_target: torch.Tensor
"""Feedforward effort (torques or forces)."""
pos: torch.Tensor
"""Current positions (joint positions, tendon lengths, or site positions)."""
vel: torch.Tensor
"""Current velocities (joint velocities, tendon velocities, or site velocities)."""
def delay_command(
buffer: DelayBuffer,
position_target: torch.Tensor,
velocity_target: torch.Tensor,
effort_target: torch.Tensor,
) -> tuple[torch.Tensor, torch.Tensor, torch.Tensor]:
"""Delay the three command-target fields through one shared channel.
Every target the policy issues (position, velocity, effort) travels the same
command bus and incurs the same latency, so they are stacked and delayed
together by a single DelayBuffer. Feedback (pos, vel) is never delayed. Shared
by Actuator.apply_delay and the fused group so the command-delay semantics
live in exactly one place.
Args:
buffer: The delay buffer to append to and read the delayed command from.
position_target: Position targets, shape (num_envs, num_targets).
velocity_target: Velocity targets, shape (num_envs, num_targets).
effort_target: Feedforward effort targets, shape (num_envs, num_targets).
Returns:
The delayed (position_target, velocity_target, effort_target) tuple.
"""
stacked = torch.stack((position_target, velocity_target, effort_target), dim=-1)
buffer.append(stacked)
delayed = buffer.compute()
return delayed[..., 0], delayed[..., 1], delayed[..., 2]
class Actuator(ABC, Generic[ActuatorCfgT]):
"""Base actuator interface."""
# Fusion contract. An actuator whose control output is a stateless function of
# per-target parameters and the command implements control_law over the
# parameter tensors named in param_names, and applies it through the shared
# compute (see IdealPdActuator). Such actuators are fused automatically (see
# mjlab.actuator.fused_group). Actuators with a custom compute (built-ins, XML,
# learned networks) are not fused; they need no opt-out flag, since overriding
# compute is itself the signal.
param_names: ClassVar[tuple[str, ...]] = ()
@staticmethod
def control_law(params: dict[str, torch.Tensor], cmd: ActuatorCmd) -> torch.Tensor:
"""Stateless control law over per-target parameters.
params maps each name in param_names to its tensor of shape
(num_envs, num_targets); for a fused group these are the concatenated
tensors. Returns the control signal of the same trailing shape. Implemented
by stateless-law actuator types.
"""
raise NotImplementedError
def __init__(
self,
cfg: ActuatorCfgT,
entity: Entity,
target_ids: list[int],
target_names: list[str],
) -> None:
self.cfg = cfg
self.entity = entity
self._target_ids_list = target_ids
self._target_names = target_names
self._target_ids: torch.Tensor | None = None
self._ctrl_ids: torch.Tensor | None = None
self._global_ctrl_ids: torch.Tensor | None = None
self._mjs_actuators: list[mujoco.MjsActuator] = []
self._site_zeros: torch.Tensor | None = None
self._delay_buffer: DelayBuffer | None = None
@property
def has_delay(self) -> bool:
"""Whether this actuator has delay configured."""
return self.cfg.delay_max_lag > 0
@property
def target_ids(self) -> torch.Tensor:
"""Local indices of targets controlled by this actuator."""
assert self._target_ids is not None
return self._target_ids
@property
def target_names(self) -> list[str]:
"""Names of targets controlled by this actuator."""
return self._target_names
@property
def transmission_type(self) -> TransmissionType:
"""Transmission type of this actuator."""
return self.cfg.transmission_type
@property
def ctrl_ids(self) -> torch.Tensor:
"""Local indices of control inputs within the entity."""
assert self._ctrl_ids is not None
return self._ctrl_ids
@property
def global_ctrl_ids(self) -> torch.Tensor:
"""Global indices of control inputs in the MuJoCo model."""
assert self._global_ctrl_ids is not None
return self._global_ctrl_ids
@abstractmethod
def edit_spec(self, spec: mujoco.MjSpec, target_names: list[str]) -> None:
"""Edit the MjSpec to add actuators.
This is called during entity construction, before the model is compiled.
Args:
spec: The entity's MjSpec to edit.
target_names: Names of targets (joints, tendons, or sites) as they
appear in the spec. When the entity's ``spec_fn`` uses internal
``MjSpec.attach(prefix=...)``, these will include the prefix
(e.g., ``"left/elbow"`` rather than ``"elbow"``).
"""
raise NotImplementedError
def initialize(
self,
mj_model: mujoco.MjModel,
model: mjwarp.Model,
data: mjwarp.Data,
device: str,
) -> None:
"""Initialize the actuator after model compilation.
This is called after the MjSpec is compiled into an MjModel.
Args:
mj_model: The compiled MuJoCo model.
model: The compiled mjwarp model.
data: The mjwarp data arrays.
device: Device for tensor operations (e.g., "cuda", "cpu").
"""
del mj_model, model # Unused.
self._target_ids = torch.tensor(
self._target_ids_list, dtype=torch.long, device=device
)
global_ctrl_ids_list = [act.id for act in self._mjs_actuators]
self._global_ctrl_ids = torch.tensor(
global_ctrl_ids_list, dtype=torch.long, device=device
)
entity_ctrl_ids = self.entity.indexing.ctrl_ids
global_to_local = {gid.item(): i for i, gid in enumerate(entity_ctrl_ids)}
self._ctrl_ids = torch.tensor(
[global_to_local[gid] for gid in global_ctrl_ids_list],
dtype=torch.long,
device=device,
)
# Pre-allocate zeros for SITE transmission type to avoid repeated allocations.
if self.transmission_type == TransmissionType.SITE:
nenvs = data.nworld
ntargets = len(self._target_ids_list)
self._site_zeros = torch.zeros((nenvs, ntargets), device=device)
self._init_delay_buffer(data.nworld, device)
def _init_delay_buffer(self, num_envs: int, device: str) -> None:
"""Create delay buffer. Called during initialize()."""
if not self.has_delay:
return
self._delay_buffer = DelayBuffer(
min_lag=self.cfg.delay_min_lag,
max_lag=self.cfg.delay_max_lag,
batch_size=num_envs,
device=device,
hold_prob=self.cfg.delay_hold_prob,
update_period=self.cfg.delay_update_period,
per_env_phase=self.cfg.delay_per_env_phase,
)
def apply_delay(self, cmd: ActuatorCmd) -> ActuatorCmd:
"""Delay all command targets with one shared lag. No-op without delay.
Every target the policy issues (position, velocity, effort) travels the same
command channel and experiences the same latency, so they are stacked and
delayed together. Feedback fields (``pos``, ``vel``) are never delayed.
"""
if self._delay_buffer is None:
return cmd
position_target, velocity_target, effort_target = delay_command(
self._delay_buffer,
cmd.position_target,
cmd.velocity_target,
cmd.effort_target,
)
return dataclasses.replace(
cmd,
position_target=position_target,
velocity_target=velocity_target,
effort_target=effort_target,
)
def set_lags(
self,
lags: torch.Tensor,
env_ids: torch.Tensor | slice | None = None,
) -> None:
"""Set delay lag values for specified environments.
Built-in actuators with the same delay config share a fused delay
buffer for performance. Calling ``set_lags`` on any one of them
affects the entire fused group.
Args:
lags: Lag values in physics timesteps. Shape: (num_env_ids,) or scalar.
env_ids: Environment indices to set. If None, sets all environments.
"""
if self._delay_buffer is not None:
self._delay_buffer.set_lags(lags, env_ids)
def get_command(self, data: EntityData) -> ActuatorCmd:
"""Extract command data for this actuator from entity data.
Args:
data: The entity data containing all state and target information.
Returns:
ActuatorCmd with appropriate data based on transmission type.
"""
if self.transmission_type == TransmissionType.JOINT:
return ActuatorCmd(
position_target=data.joint_pos_target[:, self.target_ids],
velocity_target=data.joint_vel_target[:, self.target_ids],
effort_target=data.joint_effort_target[:, self.target_ids],
pos=data.joint_pos[:, self.target_ids],
vel=data.joint_vel[:, self.target_ids],
)
elif self.transmission_type == TransmissionType.TENDON:
return ActuatorCmd(
position_target=data.tendon_len_target[:, self.target_ids],
velocity_target=data.tendon_vel_target[:, self.target_ids],
effort_target=data.tendon_effort_target[:, self.target_ids],
pos=data.tendon_len[:, self.target_ids],
vel=data.tendon_vel[:, self.target_ids],
)
elif self.transmission_type == TransmissionType.SITE:
assert self._site_zeros is not None
return ActuatorCmd(
position_target=self._site_zeros,
velocity_target=self._site_zeros,
effort_target=data.site_effort_target[:, self.target_ids],
pos=self._site_zeros,
vel=self._site_zeros,
)
else:
raise ValueError(f"Unknown transmission type: {self.transmission_type}")
@abstractmethod
def compute(self, cmd: ActuatorCmd) -> torch.Tensor:
"""Compute low-level actuator control signal from high-level commands.
Args:
cmd: High-level actuator command.
Returns:
Control signal tensor of shape (num_envs, num_actuators).
"""
raise NotImplementedError
# Optional methods.
def reset(self, env_ids: torch.Tensor | slice | None = None) -> None:
"""Reset actuator state for specified environments.
Resets delay buffers if present. Subclasses that override this should
call ``super().reset(env_ids)``.
Args:
env_ids: Environment indices to reset. If None, reset all environments.
"""
if self._delay_buffer is not None:
self._delay_buffer.reset(env_ids)
def update(self, dt: float) -> None:
"""Update actuator state after a simulation step.
Base implementation does nothing. Override in subclasses that need
per-step updates.
Args:
dt: Time step in seconds.
"""
del dt # Unused.
+631
View File
@@ -0,0 +1,631 @@
"""MuJoCo built-in actuators.
This module provides actuators that use MuJoCo's native actuator implementations,
created programmatically via the MjSpec API.
"""
from __future__ import annotations
from dataclasses import dataclass
from enum import IntEnum
from typing import TYPE_CHECKING
import mujoco
import numpy as np
import torch
from mjlab.actuator.actuator import (
Actuator,
ActuatorCfg,
ActuatorCmd,
TransmissionType,
)
from mjlab.utils.spec import (
apply_target_overrides,
create_motor_actuator,
create_muscle_actuator,
create_position_actuator,
create_velocity_actuator,
)
if TYPE_CHECKING:
from mjlab.entity import Entity
@dataclass(kw_only=True)
class BuiltinPositionActuatorCfg(ActuatorCfg):
"""Configuration for MuJoCo built-in position actuator.
Under the hood, this creates a <position> actuator for each target and sets the
stiffness, damping and effort limits accordingly. If armature or frictionloss are
set, they override the corresponding joint/tendon properties from XML.
"""
stiffness: float
"""PD proportional gain."""
damping: float
"""PD derivative gain."""
effort_limit: float | None = None
"""Maximum actuator force/torque. If None, no limit is applied."""
def __post_init__(self) -> None:
super().__post_init__()
if self.transmission_type == TransmissionType.SITE:
raise ValueError(
"BuiltinPositionActuatorCfg does not support SITE transmission. "
"Use BuiltinMotorActuatorCfg for site transmission."
)
def build(
self, entity: Entity, target_ids: list[int], target_names: list[str]
) -> BuiltinPositionActuator:
return BuiltinPositionActuator(self, entity, target_ids, target_names)
class BuiltinPositionActuator(Actuator[BuiltinPositionActuatorCfg]):
"""MuJoCo built-in position actuator."""
def __init__(
self,
cfg: BuiltinPositionActuatorCfg,
entity: Entity,
target_ids: list[int],
target_names: list[str],
) -> None:
super().__init__(cfg, entity, target_ids, target_names)
def edit_spec(self, spec: mujoco.MjSpec, target_names: list[str]) -> None:
# Add <position> actuator to spec, one per target.
for target_name in target_names:
actuator = create_position_actuator(
spec,
target_name,
stiffness=self.cfg.stiffness,
damping=self.cfg.damping,
effort_limit=self.cfg.effort_limit,
armature=self.cfg.armature,
frictionloss=self.cfg.frictionloss,
viscous_damping=self.cfg.viscous_damping,
transmission_type=self.cfg.transmission_type,
)
self._mjs_actuators.append(actuator)
def compute(self, cmd: ActuatorCmd) -> torch.Tensor:
return cmd.position_target
@dataclass(kw_only=True)
class BuiltinPdActuatorCfg(ActuatorCfg):
"""Implicit-integration version of IdealPdActuator.
Both consume a position target and a velocity target with kp/kd gains. The
difference is in how the PD is delivered to MuJoCo: IdealPdActuator computes
the PD force in Python and feeds it to a ``<motor>`` element, which MuJoCo
sees as an opaque external force. This actuator expresses the PD as native
MuJoCo elements (a ``<position>`` carrying kp, a ``<velocity>`` carrying kd),
so the implicit and implicitfast integrators include the kp/kd derivatives
in their velocity update. That makes the actuator numerically stable at
gain/timestep combinations where explicit Python PD would diverge, which
matters when you want to run a real motor's stiff on-board PD gains in sim.
"""
stiffness: float
"""Proportional gain (kp)."""
damping: float
"""Derivative gain (kd)."""
effort_limit: float | None = None
"""Maximum total torque applied to the joint or tendon. Enforced as a
sum-clamp on the two PD terms via jnt_actfrcrange (JOINT) or
tendon_actfrcrange (TENDON). None leaves the limit unset."""
def __post_init__(self) -> None:
super().__post_init__()
if self.transmission_type == TransmissionType.SITE:
raise ValueError(
"BuiltinPdActuatorCfg does not support SITE transmission. "
"Use BuiltinMotorActuatorCfg for site transmission."
)
def build(
self, entity: Entity, target_ids: list[int], target_names: list[str]
) -> BuiltinPdActuator:
return BuiltinPdActuator(self, entity, target_ids, target_names)
class BuiltinPdActuator(Actuator[BuiltinPdActuatorCfg]):
"""MuJoCo native PD: paired <position> + <velocity> elements per target."""
def __init__(
self,
cfg: BuiltinPdActuatorCfg,
entity: Entity,
target_ids: list[int],
target_names: list[str],
) -> None:
super().__init__(cfg, entity, target_ids, target_names)
@property
def num_targets(self) -> int:
"""Number of targets. ``ctrl_ids`` is laid out as ``[pos..., vel...]``,
each block of length ``num_targets``."""
return len(self._target_ids_list)
def edit_spec(self, spec: mujoco.MjSpec, target_names: list[str]) -> None:
# Position elements first, then velocity elements, so ctrl_ids is laid out
# as [pos_0..pos_{N-1}, vel_0..vel_{N-1}].
for target_name in target_names:
pos_act = create_position_actuator(
spec,
target_name,
actuator_name=f"{target_name}_pd_pos",
stiffness=self.cfg.stiffness,
damping=0.0, # damping lives on the <velocity> element.
armature=self.cfg.armature,
frictionloss=self.cfg.frictionloss,
viscous_damping=self.cfg.viscous_damping,
transmission_type=self.cfg.transmission_type,
)
self._mjs_actuators.append(pos_act)
for target_name in target_names:
vel_act = create_velocity_actuator(
spec,
target_name,
actuator_name=f"{target_name}_pd_vel",
damping=self.cfg.damping,
transmission_type=self.cfg.transmission_type,
)
self._mjs_actuators.append(vel_act)
# Effort limit: sum-clamp on the joint/tendon, not on each element.
if self.cfg.effort_limit is not None:
lim = self.cfg.effort_limit
for target_name in target_names:
if self.cfg.transmission_type == TransmissionType.JOINT:
target = spec.joint(target_name)
else:
target = spec.tendon(target_name)
target.actfrclimited = mujoco.mjtLimited.mjLIMITED_TRUE
target.actfrcrange[:] = np.array([-lim, lim])
def compute(self, cmd: ActuatorCmd) -> torch.Tensor:
return torch.cat((cmd.position_target, cmd.velocity_target), dim=1)
@dataclass(kw_only=True)
class BuiltinMotorActuatorCfg(ActuatorCfg):
"""Configuration for MuJoCo built-in motor actuator.
Under the hood, this creates a <motor> actuator for each target and sets its effort
limit and gear ratio accordingly. If armature or frictionloss are set, they override
the corresponding joint/tendon properties from XML.
"""
effort_limit: float
"""Maximum actuator effort."""
gear: float = 1.0
"""Actuator gear ratio."""
def build(
self, entity: Entity, target_ids: list[int], target_names: list[str]
) -> BuiltinMotorActuator:
return BuiltinMotorActuator(self, entity, target_ids, target_names)
class BuiltinMotorActuator(Actuator[BuiltinMotorActuatorCfg]):
"""MuJoCo built-in motor actuator."""
def __init__(
self,
cfg: BuiltinMotorActuatorCfg,
entity: Entity,
target_ids: list[int],
target_names: list[str],
) -> None:
super().__init__(cfg, entity, target_ids, target_names)
def edit_spec(self, spec: mujoco.MjSpec, target_names: list[str]) -> None:
# Add <motor> actuator to spec, one per target.
for target_name in target_names:
actuator = create_motor_actuator(
spec,
target_name,
effort_limit=self.cfg.effort_limit,
gear=self.cfg.gear,
armature=self.cfg.armature,
frictionloss=self.cfg.frictionloss,
viscous_damping=self.cfg.viscous_damping,
transmission_type=self.cfg.transmission_type,
)
self._mjs_actuators.append(actuator)
def compute(self, cmd: ActuatorCmd) -> torch.Tensor:
return cmd.effort_target
def _or_zeros(t: tuple[float, ...] | None, n: int) -> list[float]:
return list(t) if t is not None else [0.0] * n
class DcMotorInputMode(IntEnum):
"""What the ``ctrl`` signal of a ``<dcmotor>`` represents.
Values match mjtCtrlInput, consumed by mjs_setToDCMotor and stored in
actuator_ctrlspec.
"""
VOLTAGE = mujoco.mjtCtrlInput.mjINPUT_VOLTAGE.value
POSITION = mujoco.mjtCtrlInput.mjINPUT_POS.value
VELOCITY = mujoco.mjtCtrlInput.mjINPUT_VEL.value
@dataclass(frozen=True)
class DcMotorDatasheetParams:
"""Datasheet characterization of a DC motor."""
nominal_voltage: float
"""Nominal (rated) voltage V_n [V]."""
stall_torque: float
"""Stall torque tau_stall at V_n [N*m]."""
no_load_speed: float
"""No-load angular velocity omega_no_load at V_n [rad/s]."""
def _pack(self) -> tuple[list[float], float, list[float]]:
"""Returns (motorconst, resistance, nominal) for set_to_dcmotor."""
return (
[0.0, 0.0],
0.0,
[self.nominal_voltage, self.stall_torque, self.no_load_speed],
)
@dataclass(frozen=True)
class DcMotorPhysicalParams:
"""Physical characterization of a DC motor."""
kt: float
"""Torque constant [N*m/A]."""
ke: float
"""Back-EMF constant [V*s/rad]."""
resistance: float
"""Terminal resistance R [Ohm]."""
def _pack(self) -> tuple[list[float], float, list[float]]:
"""Returns (motorconst, resistance, nominal) for set_to_dcmotor."""
return [self.kt, self.ke], self.resistance, [0.0, 0.0, 0.0]
@dataclass(kw_only=True)
class BuiltinDcMotorActuatorCfg(ActuatorCfg):
"""Native MuJoCo ``<dcmotor>`` wrapper.
Models a DC motor: torque is derived from voltage via the motor constant K and
back-EMF, tau = K * (V - K * omega) / R. The back-EMF term lives in biasprm, so
MuJoCo's implicit / implicitfast integrators pick up its velocity derivative as
effective damping.
Three input modes select what ctrl carries:
* VOLTAGE: ctrl is the drive voltage. cmd.effort_target carries volts, not torque.
* POSITION / VELOCITY: an internal PID closes on the setpoint with torque-space
gains, stiffness on the position error and damping on the velocity error. The
driver compensates back-EMF, so the commanded torque is delivered until the
drive voltage reaches Vmax.
Motor characterization: pass either DcMotorDatasheetParams or DcMotorPhysicalParams
as motor_params. mjs_setToDCMotor derives K and R (including the viscous-damping
correction) and packs the generic gainprm / biasprm / dynprm slots.
Optional extensions, off by default: integral_gain / integral_limit, slew_rate,
inductance / electrical_time_constant, thermal, lugre, cogging.
dr.pd_gains randomizes only kp and kd; for DR over the extensions, write directly to
actuator_gainprm or actuator_dynprm.
"""
motor_params: DcMotorDatasheetParams | DcMotorPhysicalParams
"""Motor characterization. Datasheet form: (V_n, tau_stall, omega_no_load).
Physical form: (Kt, Ke, R)."""
mode: DcMotorInputMode = DcMotorInputMode.POSITION
"""ctrl input semantics. See class docstring."""
stiffness: float = 0.0
"""PID gain kp on the position error, in torque space. Required in POSITION
mode; must be 0 otherwise."""
damping: float = 0.0
"""PID gain kd on the velocity error, in torque space. Required in VELOCITY
mode, optional in POSITION mode; must be 0 in VOLTAGE mode."""
voltage_limit: float = 0.0
"""Max drive voltage Vmax. Required in POSITION / VELOCITY mode (clamps the
PID's drive voltage); must be 0 in VOLTAGE mode."""
integral_gain: float = 0.0
"""PID integral gain ki on the accumulated position error. POSITION mode
only."""
integral_limit: float = 0.0
"""Anti-windup clamp Imax on the integrator state. 0 disables (the
integrator can run away)."""
slew_rate: float = 0.0
"""Max rate of change of the setpoint per second. Must be 0 in VOLTAGE mode.
0 disables."""
effort_limit: float | None = None
"""Continuous torque cap [N*m]. Sets actuator_forcerange. None leaves the
per-element forcerange unset."""
gear: float = 1.0
"""Mechanical gear ratio."""
inductance: float = 0.0
"""Winding inductance L [H]. Enables first-order electrical dynamics on the
motor current. MuJoCo internally uses te = L / R; pass
electrical_time_constant directly to skip the divide. 0 disables."""
electrical_time_constant: float = 0.0
"""Alternative to inductance: specify te [s] directly. Ignored if
inductance > 0. 0 disables."""
thermal: tuple[float, float, float, float, float, float] | None = None
"""Thermal model (R_thermal, C_thermal, tau_thermal, alpha, T0, T_ambient).
See MuJoCo's ``<dcmotor thermal=...>`` reference for units and which of the
first three may be underspecified. Effective resistance becomes
R * (1 + alpha * (T + T_ambient - T0)). None disables."""
cogging: tuple[float, float, float] | None = None
"""Cogging torque (amplitude, periodicity, phase) in (N*m, cycles per unit
length, rad). Models magnetic torque ripple from rotor-stator interaction;
at joint angle q the contribution is amplitude * sin(periodicity * q + phase).
Added *after* effort_limit is enforced, matching MuJoCo's physical model:
effort_limit bounds the electromagnetic torque (the current limit), not the
mechanical torque. Total joint torque can exceed effort_limit by up to
amplitude. None disables."""
lugre: tuple[float, float, float, float, float] | None = None
"""LuGre friction (sigma0, sigma1, F_Coulomb, F_Stribeck, v_Stribeck).
Stick-slip friction with bristle-deflection state. Subtracted from joint
torque after the effort_limit clamp (mechanical, like cogging). None
disables."""
def __post_init__(self) -> None:
super().__post_init__()
if self.transmission_type == TransmissionType.SITE:
raise ValueError(
"BuiltinDcMotorActuatorCfg does not support SITE transmission. "
"Use BuiltinMotorActuatorCfg for site transmission."
)
if self.mode in (DcMotorInputMode.POSITION, DcMotorInputMode.VELOCITY):
if self.mode == DcMotorInputMode.POSITION and self.stiffness <= 0.0:
raise ValueError("POSITION mode requires stiffness > 0.")
if self.mode == DcMotorInputMode.VELOCITY and (
self.stiffness != 0.0 or self.damping <= 0.0
):
raise ValueError("VELOCITY mode requires damping > 0 and stiffness = 0.")
if self.voltage_limit <= 0.0:
raise ValueError(f"{self.mode.name} mode requires voltage_limit > 0.")
elif any((self.stiffness, self.damping, self.voltage_limit, self.slew_rate)):
raise ValueError(
"stiffness, damping, voltage_limit, and slew_rate are unused in VOLTAGE mode."
)
if self.integral_gain != 0.0 and self.mode != DcMotorInputMode.POSITION:
raise ValueError("integral_gain requires POSITION mode.")
for name in (
"integral_gain",
"integral_limit",
"slew_rate",
"inductance",
"electrical_time_constant",
):
if getattr(self, name) < 0.0:
raise ValueError(f"{name} must be non-negative.")
def build(
self, entity: Entity, target_ids: list[int], target_names: list[str]
) -> BuiltinDcMotorActuator:
return BuiltinDcMotorActuator(self, entity, target_ids, target_names)
class BuiltinDcMotorActuator(Actuator[BuiltinDcMotorActuatorCfg]):
"""MuJoCo native ``<dcmotor>``: one actuator per target."""
def edit_spec(self, spec: mujoco.MjSpec, target_names: list[str]) -> None:
cfg = self.cfg
motorconst, resistance, nominal = cfg.motor_params._pack()
saturation = (
[cfg.effort_limit, 0.0, 0.0] if cfg.effort_limit is not None else [0.0] * 3
)
controller = [
cfg.stiffness, # kp
cfg.integral_gain, # ki
cfg.damping, # kd
cfg.slew_rate, # slewmax
cfg.integral_limit, # Imax (anti-windup)
cfg.voltage_limit, # v_max
]
# SITE is rejected in __post_init__, so only JOINT and TENDON remain.
trntype = (
mujoco.mjtTrn.mjTRN_JOINT
if cfg.transmission_type == TransmissionType.JOINT
else mujoco.mjtTrn.mjTRN_TENDON
)
for target_name in target_names:
actuator = spec.add_actuator(name=target_name, target=target_name)
actuator.trntype = trntype
actuator.gear[0] = cfg.gear
actuator.set_to_dcmotor(
motorconst=motorconst,
resistance=resistance,
nominal=nominal,
saturation=saturation,
controller=controller,
cogging=_or_zeros(cfg.cogging, 3),
inductance=[cfg.inductance, cfg.electrical_time_constant],
thermal=_or_zeros(cfg.thermal, 6),
lugre=_or_zeros(cfg.lugre, 5),
ctrlspec=cfg.mode,
)
apply_target_overrides(
spec,
target_name,
cfg.transmission_type,
armature=cfg.armature,
frictionloss=cfg.frictionloss,
viscous_damping=cfg.viscous_damping,
)
self._mjs_actuators.append(actuator)
def compute(self, cmd: ActuatorCmd) -> torch.Tensor:
if self.cfg.mode == DcMotorInputMode.POSITION:
return cmd.position_target
if self.cfg.mode == DcMotorInputMode.VELOCITY:
return cmd.velocity_target
# voltage mode: ctrl is the drive voltage carried in effort_target.
return cmd.effort_target
@dataclass(kw_only=True)
class BuiltinVelocityActuatorCfg(ActuatorCfg):
"""Configuration for MuJoCo built-in velocity actuator.
Under the hood, this creates a <velocity> actuator for each target and sets the
damping gain. If armature or frictionloss are set, they override the corresponding
joint/tendon properties from XML.
"""
damping: float
"""Damping gain."""
effort_limit: float | None = None
"""Maximum actuator force/torque. If None, no limit is applied."""
def __post_init__(self) -> None:
super().__post_init__()
if self.transmission_type == TransmissionType.SITE:
raise ValueError(
"BuiltinVelocityActuatorCfg does not support SITE transmission. "
"Use BuiltinMotorActuatorCfg for site transmission."
)
def build(
self, entity: Entity, target_ids: list[int], target_names: list[str]
) -> BuiltinVelocityActuator:
return BuiltinVelocityActuator(self, entity, target_ids, target_names)
class BuiltinVelocityActuator(Actuator[BuiltinVelocityActuatorCfg]):
"""MuJoCo built-in velocity actuator."""
def __init__(
self,
cfg: BuiltinVelocityActuatorCfg,
entity: Entity,
target_ids: list[int],
target_names: list[str],
) -> None:
super().__init__(cfg, entity, target_ids, target_names)
def edit_spec(self, spec: mujoco.MjSpec, target_names: list[str]) -> None:
# Add <velocity> actuator to spec, one per target.
for target_name in target_names:
actuator = create_velocity_actuator(
spec,
target_name,
damping=self.cfg.damping,
effort_limit=self.cfg.effort_limit,
armature=self.cfg.armature,
frictionloss=self.cfg.frictionloss,
viscous_damping=self.cfg.viscous_damping,
transmission_type=self.cfg.transmission_type,
)
self._mjs_actuators.append(actuator)
def compute(self, cmd: ActuatorCmd) -> torch.Tensor:
return cmd.velocity_target
@dataclass(kw_only=True)
class BuiltinMuscleActuatorCfg(ActuatorCfg):
"""Configuration for MuJoCo built-in muscle actuator."""
length_range: tuple[float, float] = (0.0, 0.0)
"""Length range for muscle actuator."""
gear: float = 1.0
"""Gear ratio."""
timeconst: tuple[float, float] = (0.01, 0.04)
"""Activation and deactivation time constants."""
tausmooth: float = 0.0
"""Smoothing time constant."""
range: tuple[float, float] = (0.75, 1.05)
"""Operating range of normalized muscle length."""
force: float = -1.0
"""Peak force (if -1, defaults to scale * FLV)."""
scale: float = 200.0
"""Force scaling factor."""
lmin: float = 0.5
"""Minimum normalized muscle length."""
lmax: float = 1.6
"""Maximum normalized muscle length."""
vmax: float = 1.5
"""Maximum normalized muscle velocity."""
fpmax: float = 1.3
"""Passive force at lmax."""
fvmax: float = 1.2
"""Active force at -vmax."""
def __post_init__(self) -> None:
super().__post_init__()
if self.transmission_type == TransmissionType.SITE:
raise ValueError(
"BuiltinMuscleActuatorCfg does not support SITE transmission. "
"Use BuiltinMotorActuatorCfg for site transmission."
)
def build(
self, entity: Entity, target_ids: list[int], target_names: list[str]
) -> BuiltinMuscleActuator:
return BuiltinMuscleActuator(self, entity, target_ids, target_names)
class BuiltinMuscleActuator(Actuator[BuiltinMuscleActuatorCfg]):
"""MuJoCo built-in muscle actuator."""
def __init__(
self,
cfg: BuiltinMuscleActuatorCfg,
entity: Entity,
target_ids: list[int],
target_names: list[str],
) -> None:
super().__init__(cfg, entity, target_ids, target_names)
def edit_spec(self, spec: mujoco.MjSpec, target_names: list[str]) -> None:
# Add <muscle> actuator to spec, one per target.
for target_name in target_names:
actuator = create_muscle_actuator(
spec,
target_name,
length_range=self.cfg.length_range,
gear=self.cfg.gear,
timeconst=self.cfg.timeconst,
tausmooth=self.cfg.tausmooth,
range=self.cfg.range,
force=self.cfg.force,
scale=self.cfg.scale,
lmin=self.cfg.lmin,
lmax=self.cfg.lmax,
vmax=self.cfg.vmax,
fpmax=self.cfg.fpmax,
fvmax=self.cfg.fvmax,
transmission_type=self.cfg.transmission_type,
)
self._mjs_actuators.append(actuator)
def compute(self, cmd: ActuatorCmd) -> torch.Tensor:
return cmd.effort_target
+192
View File
@@ -0,0 +1,192 @@
"""Batch ctrl writes for builtin actuators.
Builtin actuators have identity compute(): they just pass the target through to ctrl.
Instead of looping over each actuator individually, this module reads the target tensor
and writes ctrl in one batched operation per actuator type. Delayed builtins go through
a shared DelayBuffer first.
"""
from __future__ import annotations
from dataclasses import dataclass, field
from typing import TYPE_CHECKING
import torch
from mjlab.actuator.actuator import TransmissionType
from mjlab.actuator.builtin_actuator import (
BuiltinMotorActuator,
BuiltinMuscleActuator,
BuiltinPositionActuator,
BuiltinVelocityActuator,
)
from mjlab.utils.buffers import DelayBuffer
if TYPE_CHECKING:
from mjlab.actuator.actuator import Actuator
from mjlab.entity.data import EntityData
BuiltinActuatorType = (
BuiltinMotorActuator
| BuiltinMuscleActuator
| BuiltinPositionActuator
| BuiltinVelocityActuator
)
# Maps (actuator_type, transmission_type) to EntityData target tensor attribute name.
_TARGET_TENSOR_MAP: dict[tuple[type[BuiltinActuatorType], TransmissionType], str] = {
(BuiltinPositionActuator, TransmissionType.JOINT): "joint_pos_target",
(BuiltinVelocityActuator, TransmissionType.JOINT): "joint_vel_target",
(BuiltinMotorActuator, TransmissionType.JOINT): "joint_effort_target",
(BuiltinPositionActuator, TransmissionType.TENDON): "tendon_len_target",
(BuiltinVelocityActuator, TransmissionType.TENDON): "tendon_vel_target",
(BuiltinMotorActuator, TransmissionType.TENDON): "tendon_effort_target",
(BuiltinMotorActuator, TransmissionType.SITE): "site_effort_target",
(BuiltinMuscleActuator, TransmissionType.JOINT): "joint_effort_target",
(BuiltinMuscleActuator, TransmissionType.TENDON): "tendon_effort_target",
}
@dataclass
class _FusedDelayGroup:
"""A fused group of delayed builtin actuators sharing delay config."""
target_attr: str
target_ids: torch.Tensor
ctrl_ids: torch.Tensor
min_lag: int
max_lag: int
hold_prob: float
update_period: int
per_env_phase: bool
absorbed_actuators: list[Actuator] = field(default_factory=list)
delay_buffer: DelayBuffer | None = field(default=None, init=False)
@dataclass
class BuiltinActuatorGroup:
"""Groups builtin actuators for batch processing."""
_index_groups: dict[
tuple[type[BuiltinActuatorType], TransmissionType],
tuple[torch.Tensor, torch.Tensor],
]
_delayed_groups: list[_FusedDelayGroup]
@staticmethod
def process(
actuators: list[Actuator],
) -> tuple[BuiltinActuatorGroup, tuple[Actuator, ...]]:
"""Classify actuators into builtin groups and custom actuators.
Non-delayed builtins go into direct index groups. Delayed builtins with matching
delay config are fused into shared delay buffer groups. Everything else is returned
as custom actuators.
Args:
actuators: List of initialized actuators to process.
Returns:
A tuple of (builtin group, custom actuators).
"""
builtin_groups: dict[
tuple[type[BuiltinActuatorType], TransmissionType], list[Actuator]
] = {}
delayed_grouped: dict[tuple, list[Actuator]] = {}
custom_actuators: list[Actuator] = []
for act in actuators:
if not isinstance(act, BuiltinActuatorType):
custom_actuators.append(act)
continue
if not act.has_delay:
key: tuple[type[BuiltinActuatorType], TransmissionType] = (
type(act),
act.cfg.transmission_type,
)
builtin_groups.setdefault(key, []).append(act)
continue
# Delayed builtin: fuse into shared delay buffer group.
delay_key = (
type(act),
act.cfg.transmission_type,
act.cfg.delay_min_lag,
act.cfg.delay_max_lag,
act.cfg.delay_hold_prob,
act.cfg.delay_update_period,
act.cfg.delay_per_env_phase,
)
delayed_grouped.setdefault(delay_key, []).append(act)
# Build non-delayed index groups.
index_groups: dict[
tuple[type[BuiltinActuatorType], TransmissionType],
tuple[torch.Tensor, torch.Tensor],
] = {
key: (
torch.cat([a.target_ids for a in acts], dim=0),
torch.cat([a.ctrl_ids for a in acts], dim=0),
)
for key, acts in builtin_groups.items()
}
# Build delayed fused groups.
delayed_groups: list[_FusedDelayGroup] = []
for (base_type, transmission_type, *_), acts in delayed_grouped.items():
target_attr = _TARGET_TENSOR_MAP[(base_type, transmission_type)]
target_ids = torch.cat([a.target_ids for a in acts], dim=0)
ctrl_ids = torch.cat([a.ctrl_ids for a in acts], dim=0)
cfg = acts[0].cfg
delayed_groups.append(
_FusedDelayGroup(
target_attr=target_attr,
target_ids=target_ids,
ctrl_ids=ctrl_ids,
min_lag=cfg.delay_min_lag,
max_lag=cfg.delay_max_lag,
hold_prob=cfg.delay_hold_prob,
update_period=cfg.delay_update_period,
per_env_phase=cfg.delay_per_env_phase,
absorbed_actuators=list(acts),
)
)
return BuiltinActuatorGroup(index_groups, delayed_groups), tuple(custom_actuators)
def initialize(self, num_envs: int, device: str) -> None:
"""Create fused delay buffers for delayed builtin groups."""
for group in self._delayed_groups:
group.delay_buffer = DelayBuffer(
min_lag=group.min_lag,
max_lag=group.max_lag,
batch_size=num_envs,
device=device,
hold_prob=group.hold_prob,
update_period=group.update_period,
per_env_phase=group.per_env_phase,
)
# Alias the fused buffer into each absorbed actuator so that per-actuator reset
# and set_lags operate on the shared buffer.
for act in group.absorbed_actuators:
act._delay_buffer = group.delay_buffer
def apply_controls(self, data: EntityData) -> None:
"""Write builtin actuator controls to simulation data."""
# Non-delayed: direct write.
for (actuator_type, transmission_type), (
target_ids,
ctrl_ids,
) in self._index_groups.items():
attr_name = _TARGET_TENSOR_MAP[(actuator_type, transmission_type)]
target_tensor = getattr(data, attr_name)
data.write_ctrl(target_tensor[:, target_ids], ctrl_ids)
# Delayed: append to buffer, compute delayed value, write.
for group in self._delayed_groups:
assert group.delay_buffer is not None
target_tensor = getattr(data, group.target_attr)
targets = target_tensor[:, group.target_ids]
group.delay_buffer.append(targets)
data.write_ctrl(group.delay_buffer.compute(), group.ctrl_ids)
+189
View File
@@ -0,0 +1,189 @@
"""DC motor actuator with velocity-based saturation model.
This module provides a DC motor actuator that implements a realistic torque-speed
curve for more accurate motor behavior simulation.
"""
from __future__ import annotations
from dataclasses import dataclass
from typing import TYPE_CHECKING, Generic, TypeVar
import mujoco
import mujoco_warp as mjwarp
import torch
from mjlab.actuator.actuator import ActuatorCmd
from mjlab.actuator.pd_actuator import IdealPdActuator, IdealPdActuatorCfg, pd_torque
if TYPE_CHECKING:
from mjlab.entity import Entity
DcMotorCfgT = TypeVar("DcMotorCfgT", bound="DcMotorActuatorCfg")
def dc_motor_clip(
effort: torch.Tensor,
saturation_effort: torch.Tensor,
velocity_limit: torch.Tensor,
force_limit: torch.Tensor,
vel: torch.Tensor,
) -> torch.Tensor:
"""Clip effort to the DC motor torque-speed curve.
Linear torque-speed curve: full saturation_effort at zero velocity falling to
zero at velocity_limit, further bounded by the continuous force_limit. Shared
by DcMotorActuator._clip_effort and the fused control law so the curve lives in
one place.
The corner velocity where the curve intersects force_limit is recomputed from
force_limit every call rather than cached, since force_limit can be
domain-randomized after initialize() (e.g. via set_effort_limit); caching it
would silently go stale.
"""
vel_at_effort_lim = velocity_limit * (1 + force_limit / saturation_effort)
vel_clipped = torch.clamp(vel, min=-vel_at_effort_lim, max=vel_at_effort_lim)
torque_speed_top = saturation_effort * (1.0 - vel_clipped / velocity_limit)
torque_speed_bottom = saturation_effort * (-1.0 - vel_clipped / velocity_limit)
max_effort = torch.clamp(torque_speed_top, max=force_limit)
min_effort = torch.clamp(torque_speed_bottom, min=-force_limit)
return torch.clamp(effort, min=min_effort, max=max_effort)
@dataclass(kw_only=True)
class DcMotorActuatorCfg(IdealPdActuatorCfg):
"""Configuration for DC motor actuator with velocity-based saturation.
This actuator implements a DC motor torque-speed curve for more realistic
actuator behavior. The motor produces maximum torque (saturation_effort) at
zero velocity and reduces linearly to zero torque at maximum velocity.
Note: effort_limit should be explicitly set to a realistic value for proper
motor modeling. Using the default (inf) will trigger a warning. Use
IdealPdActuator if unlimited torque is desired.
For a native MuJoCo ``<dcmotor>`` with back-EMF, voltage saturation, and
configurable ``Kt`` / ``Ke`` / ``R``, see ``BuiltinDcMotorActuator``.
"""
saturation_effort: float
"""Peak motor torque at zero velocity (stall torque)."""
velocity_limit: float
"""Maximum motor velocity (no-load speed)."""
def __post_init__(self) -> None:
"""Validate DC motor parameters."""
super().__post_init__()
import warnings
if self.effort_limit == float("inf"):
warnings.warn(
"effort_limit is set to inf for DcMotorActuator, which creates an "
"unrealistic motor with unlimited continuous torque. Consider setting "
"effort_limit to your motor's continuous rating (<= saturation_effort). "
"Use IdealPdActuator if you truly want unlimited torque.",
UserWarning,
stacklevel=2,
)
if self.effort_limit > self.saturation_effort:
warnings.warn(
f"effort_limit ({self.effort_limit}) exceeds saturation_effort "
f"({self.saturation_effort}). For realistic motors, continuous torque "
"should be <= peak torque.",
UserWarning,
stacklevel=2,
)
def build(
self, entity: Entity, target_ids: list[int], target_names: list[str]
) -> DcMotorActuator:
return DcMotorActuator(self, entity, target_ids, target_names)
class DcMotorActuator(IdealPdActuator[DcMotorCfgT], Generic[DcMotorCfgT]):
"""DC motor actuator with velocity-based saturation model.
This actuator extends IdealPdActuator with a realistic DC motor model
that limits torque based on current joint velocity. The model implements
a linear torque-speed curve where:
- At zero velocity: can produce full saturation_effort (stall torque)
- At max velocity: can produce zero torque
- Between: torque limit varies linearly
The continuous torque limit (effort_limit) further constrains the output.
"""
# Same stateless-law compute as IdealPd, with the torque-speed curve added to
# the law. The velocity feedback the curve needs comes straight from cmd.vel.
param_names = (
*IdealPdActuator.param_names,
"saturation_effort",
"velocity_limit_motor",
)
@staticmethod
def control_law(params: dict[str, torch.Tensor], cmd: ActuatorCmd) -> torch.Tensor:
torque = pd_torque(params["stiffness"], params["damping"], cmd)
return dc_motor_clip(
torque,
params["saturation_effort"],
params["velocity_limit_motor"],
params["force_limit"],
cmd.vel,
)
def __init__(
self,
cfg: DcMotorCfgT,
entity: Entity,
target_ids: list[int],
target_names: list[str],
) -> None:
super().__init__(cfg, entity, target_ids, target_names)
self.saturation_effort: torch.Tensor | None = None
self.velocity_limit_motor: torch.Tensor | None = None
self._joint_vel_clipped: torch.Tensor | None = None
def initialize(
self,
mj_model: mujoco.MjModel,
model: mjwarp.Model,
data: mjwarp.Data,
device: str,
) -> None:
super().initialize(mj_model, model, data, device)
num_envs = data.nworld
num_joints = len(self._target_names)
self.saturation_effort = torch.full(
(num_envs, num_joints),
self.cfg.saturation_effort,
dtype=torch.float,
device=device,
)
self.velocity_limit_motor = torch.full(
(num_envs, num_joints),
self.cfg.velocity_limit,
dtype=torch.float,
device=device,
)
self._joint_vel_clipped = torch.zeros(num_envs, num_joints, device=device)
def _clip_effort(self, effort: torch.Tensor) -> torch.Tensor:
# Retained for LearnedMlpActuator, which has its own (stateful) compute and
# reuses this torque-speed clip. DcMotorActuator itself clips inside
# control_law via dc_motor_clip.
assert self.saturation_effort is not None
assert self.velocity_limit_motor is not None
assert self.force_limit is not None
assert self._joint_vel_clipped is not None
return dc_motor_clip(
effort,
self.saturation_effort,
self.velocity_limit_motor,
self.force_limit,
self._joint_vel_clipped,
)
+223
View File
@@ -0,0 +1,223 @@
"""Batch compute for stateless-control-law actuators.
Actuators like IdealPdActuator and DcMotorActuator evaluate their control law in
PyTorch every physics substep. Done per actuator, a robot split into several
groups (e.g. one per motor type) issues that many gathers, delay-buffer updates,
control-law evaluations, and ctrl writes each substep. At small env counts (play
mode with a synchronizing viewer) this is host-launch-bound and noticeably slow,
whereas built-in actuators are already fused into one batched write.
This module brings the same fusion to actuators whose control output is a
stateless function of per-target parameters and the command. Such actuators apply
their control law through the shared compute inherited from IdealPdActuator;
keeping that compute is exactly what marks an actuator as fusable. Actuators with
a custom compute (built-ins, XML, learned networks) are skipped, with no opt-out
flag required.
Fusable actuators sharing an exact type, transmission type, and delay config are
fused into one group: their per-target parameters and indices are concatenated so
the gather, optional delay, control law, and ctrl write each happen once for the
whole group, behind a single shared DelayBuffer. SITE transmission is not fused.
"""
from __future__ import annotations
from dataclasses import dataclass, field
from typing import TYPE_CHECKING
import torch
from mjlab.actuator.actuator import ActuatorCmd, TransmissionType, delay_command
from mjlab.actuator.pd_actuator import IdealPdActuator
from mjlab.utils.buffers import DelayBuffer
if TYPE_CHECKING:
from mjlab.actuator.actuator import Actuator
from mjlab.entity.data import EntityData
# Per transmission type, the EntityData attribute names for the five fields the
# control law reads: (position_target, velocity_target, effort_target, position,
# velocity).
_FIELD_MAP: dict[TransmissionType, tuple[str, str, str, str, str]] = {
TransmissionType.JOINT: (
"joint_pos_target",
"joint_vel_target",
"joint_effort_target",
"joint_pos",
"joint_vel",
),
TransmissionType.TENDON: (
"tendon_len_target",
"tendon_vel_target",
"tendon_effort_target",
"tendon_len",
"tendon_vel",
),
}
def _is_fusable(act: Actuator) -> bool:
"""An actuator is fusable iff it keeps the shared stateless-law compute.
IdealPd and DcMotor apply their control law through that inherited compute;
anything with a custom compute (built-ins, XML, and learned actuators,
including DcMotor subclasses that override it) is not fusable. Comparing the
bound function is immune to the inheritance trap an isinstance check would hit.
"""
return (
type(act).compute is IdealPdActuator.compute
and act.cfg.transmission_type in _FIELD_MAP
)
def _require(t: torch.Tensor | None) -> torch.Tensor:
assert t is not None
return t
@dataclass
class _FusedGroup:
"""A fused group of actuators sharing type, transmission, and delay config."""
actuator_type: type[Actuator]
transmission_type: TransmissionType
target_ids: torch.Tensor
ctrl_ids: torch.Tensor
min_lag: int
max_lag: int
hold_prob: float
update_period: int
per_env_phase: bool
absorbed_actuators: list[Actuator] = field(default_factory=list)
params: dict[str, torch.Tensor] = field(default_factory=dict, init=False)
delay_buffer: DelayBuffer | None = field(default=None, init=False)
@property
def has_delay(self) -> bool:
return self.max_lag > 0
@dataclass
class FusedActuatorGroup:
"""Groups stateless-control-law actuators for batch processing."""
_groups: list[_FusedGroup]
@staticmethod
def process(
actuators: tuple[Actuator, ...] | list[Actuator],
) -> tuple[FusedActuatorGroup, tuple[Actuator, ...]]:
"""Classify actuators into fused groups and remaining custom actuators.
Fusable actuators (see _is_fusable) with JOINT or TENDON transmission are
fused by (exact type, transmission type, delay config). Everything else is
returned unchanged for the per-actuator path.
Args:
actuators: List of initialized actuators to process.
Returns:
A tuple of (fused group, remaining custom actuators).
"""
grouped: dict[tuple, list[Actuator]] = {}
remaining: list[Actuator] = []
for act in actuators:
if not _is_fusable(act):
remaining.append(act)
continue
key = (
type(act),
act.cfg.transmission_type,
act.cfg.delay_min_lag,
act.cfg.delay_max_lag,
act.cfg.delay_hold_prob,
act.cfg.delay_update_period,
act.cfg.delay_per_env_phase,
)
grouped.setdefault(key, []).append(act)
groups: list[_FusedGroup] = []
for (actuator_type, transmission_type, *_), acts in grouped.items():
cfg = acts[0].cfg
groups.append(
_FusedGroup(
actuator_type=actuator_type,
transmission_type=transmission_type,
target_ids=torch.cat([a.target_ids for a in acts], dim=0),
ctrl_ids=torch.cat([a.ctrl_ids for a in acts], dim=0),
min_lag=cfg.delay_min_lag,
max_lag=cfg.delay_max_lag,
hold_prob=cfg.delay_hold_prob,
update_period=cfg.delay_update_period,
per_env_phase=cfg.delay_per_env_phase,
absorbed_actuators=list(acts),
)
)
return FusedActuatorGroup(groups), tuple(remaining)
def initialize(self, num_envs: int, device: str) -> None:
"""Concatenate parameters and create shared delay buffers for fused groups."""
for group in self._groups:
# Concatenate each control-law parameter across absorbed actuators, then
# alias every actuator's parameter back to a slice view of the fused tensor.
#
# Invariant: after this an absorbed actuator's parameter tensors are views
# into the group tensors. Mutations must be in-place (e.g. set_gains and
# set_effort_limit assign into a slice) so they write through to what the
# group computes with. Rebinding the attribute would silently detach it; the
# only rebind site is the actuator's initialize(), which runs before fusion.
for name in group.actuator_type.param_names:
group.params[name] = torch.cat(
[_require(getattr(a, name)) for a in group.absorbed_actuators], dim=1
)
offset = 0
for act in group.absorbed_actuators:
n = act.target_ids.numel()
for name in group.actuator_type.param_names:
setattr(act, name, group.params[name][:, offset : offset + n])
offset += n
if group.has_delay:
group.delay_buffer = DelayBuffer(
min_lag=group.min_lag,
max_lag=group.max_lag,
batch_size=num_envs,
device=device,
hold_prob=group.hold_prob,
update_period=group.update_period,
per_env_phase=group.per_env_phase,
)
# Alias the shared buffer into each absorbed actuator so per-actuator
# reset and set_lags operate on it.
for act in group.absorbed_actuators:
act._delay_buffer = group.delay_buffer
def apply_controls(self, data: EntityData) -> None:
"""Compute and write fused actuator controls to simulation data."""
for group in self._groups:
pos_attr, vel_attr, eff_attr, cur_pos_attr, cur_vel_attr = _FIELD_MAP[
group.transmission_type
]
ids = group.target_ids
pos_target = getattr(data, pos_attr)[:, ids]
vel_target = getattr(data, vel_attr)[:, ids]
effort_target = getattr(data, eff_attr)[:, ids]
pos = getattr(data, cur_pos_attr)[:, ids]
vel = getattr(data, cur_vel_attr)[:, ids]
if group.delay_buffer is not None:
pos_target, vel_target, effort_target = delay_command(
group.delay_buffer, pos_target, vel_target, effort_target
)
cmd = ActuatorCmd(
position_target=pos_target,
velocity_target=vel_target,
effort_target=effort_target,
pos=pos,
vel=vel,
)
torques = group.actuator_type.control_law(group.params, cmd)
data.write_ctrl(torques, group.ctrl_ids)
+208
View File
@@ -0,0 +1,208 @@
"""Learned actuator models."""
from __future__ import annotations
from dataclasses import dataclass
from typing import TYPE_CHECKING, Literal
import mujoco
import mujoco_warp as mjwarp
import torch
from mjlab.actuator.actuator import ActuatorCmd
from mjlab.actuator.dc_actuator import DcMotorActuator, DcMotorActuatorCfg
from mjlab.utils.buffers import CircularBuffer
if TYPE_CHECKING:
from mjlab.entity import Entity
@dataclass(kw_only=True)
class LearnedMlpActuatorCfg(DcMotorActuatorCfg):
"""Configuration for MLP-based learned actuator model.
This actuator learns the mapping from joint commands and state history to
actual torque output. It's useful for capturing actuator dynamics that are
difficult to model analytically, such as delays, non-linearities, and
friction effects.
The network is trained offline using data from the real system and loaded
as a TorchScript file. The model uses a sliding window of historical joint
position errors and velocities as inputs.
"""
network_file: str
"""Path to the TorchScript file containing the trained MLP model."""
pos_scale: float
"""Scaling factor for joint position error inputs to the network."""
vel_scale: float
"""Scaling factor for joint velocity inputs to the network."""
torque_scale: float
"""Scaling factor for torque outputs from the network."""
input_order: Literal["pos_vel", "vel_pos"] = "pos_vel"
"""Order of inputs to the network.
- "pos_vel": position errors followed by velocities
- "vel_pos": velocities followed by position errors
"""
history_length: int = 3
"""Number of timesteps of history to use as network inputs.
For example, history_length=3 uses the current timestep plus the previous
2 timesteps (total of 3 frames).
"""
# Learned actuators don't use stiffness/damping from PD controller.
stiffness: float = 0.0
damping: float = 0.0
def build(
self, entity: Entity, target_ids: list[int], target_names: list[str]
) -> LearnedMlpActuator:
return LearnedMlpActuator(self, entity, target_ids, target_names)
class LearnedMlpActuator(DcMotorActuator[LearnedMlpActuatorCfg]):
"""MLP-based learned actuator with joint history.
This actuator uses a trained neural network to map from joint commands and
state history to torque outputs. The network captures complex actuator
dynamics that are difficult to model analytically.
The actuator maintains circular buffers of joint position errors and
velocities, which are used as inputs to the MLP. The network outputs are
then scaled and clipped using the DC motor limits from the parent class.
"""
def __init__(
self,
cfg: LearnedMlpActuatorCfg,
entity: Entity,
target_ids: list[int],
target_names: list[str],
) -> None:
super().__init__(cfg, entity, target_ids, target_names)
self.network: torch.jit.ScriptModule | None = None
self._pos_error_history: CircularBuffer | None = None
self._vel_history: CircularBuffer | None = None
def initialize(
self,
mj_model: mujoco.MjModel,
model: mjwarp.Model,
data: mjwarp.Data,
device: str,
) -> None:
super().initialize(mj_model, model, data, device)
# Load the trained network from TorchScript file.
self.network = torch.jit.load(self.cfg.network_file, map_location=device)
assert self.network is not None
self.network.eval()
# Create history buffers.
num_envs = data.nworld
self._pos_error_history = CircularBuffer(
max_len=self.cfg.history_length,
batch_size=num_envs,
device=device,
)
self._vel_history = CircularBuffer(
max_len=self.cfg.history_length,
batch_size=num_envs,
device=device,
)
def reset(self, env_ids: torch.Tensor | slice | None = None) -> None:
"""Reset history buffers for specified environments.
Args:
env_ids: Environment indices to reset. If None, reset all environments.
"""
super().reset(env_ids)
assert self._pos_error_history is not None
assert self._vel_history is not None
if env_ids is None:
self._pos_error_history.reset()
self._vel_history.reset()
elif isinstance(env_ids, slice):
# Convert slice to indices for CircularBuffer.
batch_size = self._pos_error_history.batch_size
indices = list(range(*env_ids.indices(batch_size)))
self._pos_error_history.reset(indices)
self._vel_history.reset(indices)
else:
self._pos_error_history.reset(env_ids)
self._vel_history.reset(env_ids)
def compute(self, cmd: ActuatorCmd) -> torch.Tensor:
"""Compute actuator torques using the learned MLP model.
Args:
cmd: High-level actuator command containing targets and current state.
Returns:
Computed torque tensor of shape (num_envs, num_joints).
"""
assert self.network is not None
assert self._pos_error_history is not None
assert self._vel_history is not None
assert self._joint_vel_clipped is not None
# Update history buffers with current state.
pos_error = cmd.position_target - cmd.pos
self._pos_error_history.append(pos_error)
self._vel_history.append(cmd.vel)
# Save velocity for DC motor clipping in parent class.
self._joint_vel_clipped[:] = cmd.vel
num_envs = cmd.pos.shape[0]
num_joints = cmd.pos.shape[1]
# Extract history from current to history_length-1 steps back.
# Each returns shape: (num_envs, num_joints).
pos_inputs = [
self._pos_error_history[lag] for lag in range(self.cfg.history_length)
]
vel_inputs = [self._vel_history[lag] for lag in range(self.cfg.history_length)]
# Stack along feature dimension: (num_envs, num_joints, history_length).
pos_stacked = torch.stack(pos_inputs, dim=2)
vel_stacked = torch.stack(vel_inputs, dim=2)
# Reshape to (num_envs * num_joints, num_history_steps) for network.
pos_flat = pos_stacked.reshape(num_envs * num_joints, -1)
vel_flat = vel_stacked.reshape(num_envs * num_joints, -1)
# Scale and concatenate inputs based on specified order.
if self.cfg.input_order == "pos_vel":
network_input = torch.cat(
[pos_flat * self.cfg.pos_scale, vel_flat * self.cfg.vel_scale], dim=1
)
elif self.cfg.input_order == "vel_pos":
network_input = torch.cat(
[vel_flat * self.cfg.vel_scale, pos_flat * self.cfg.pos_scale], dim=1
)
else:
raise ValueError(
f"Invalid input order: {self.cfg.input_order}. Must be 'pos_vel' or 'vel_pos'."
)
# Run network inference.
with torch.inference_mode():
torques_flat = self.network(network_input)
# Reshape and scale output torques.
computed_torques = torques_flat.reshape(num_envs, num_joints)
computed_torques = computed_torques * self.cfg.torque_scale
# Clip using DC motor limits from parent class.
return self._clip_effort(computed_torques)
+163
View File
@@ -0,0 +1,163 @@
"""An ideal PD control actuator."""
from __future__ import annotations
from dataclasses import dataclass
from typing import TYPE_CHECKING, Generic, TypeVar
import mujoco
import mujoco_warp as mjwarp
import torch
from mjlab.actuator.actuator import Actuator, ActuatorCfg, ActuatorCmd
from mjlab.utils.spec import create_motor_actuator
if TYPE_CHECKING:
from mjlab.entity import Entity
IdealPdCfgT = TypeVar("IdealPdCfgT", bound="IdealPdActuatorCfg")
def pd_torque(
stiffness: torch.Tensor, damping: torch.Tensor, cmd: ActuatorCmd
) -> torch.Tensor:
"""Ideal PD control torque, unclamped.
kp * (pos_target - pos) + kd * (vel_target - vel) + effort_target. Shared by
the ideal PD and DC motor control laws so the PD sum lives in one place.
"""
torque = stiffness * (cmd.position_target - cmd.pos)
torque += damping * (cmd.velocity_target - cmd.vel)
torque += cmd.effort_target
return torque
@dataclass(kw_only=True)
class IdealPdActuatorCfg(ActuatorCfg):
"""Configuration for ideal PD actuator."""
stiffness: float
"""PD stiffness (proportional gain)."""
damping: float
"""PD damping (derivative gain)."""
effort_limit: float = float("inf")
"""Maximum force/torque limit."""
def build(
self, entity: Entity, target_ids: list[int], target_names: list[str]
) -> IdealPdActuator:
return IdealPdActuator(self, entity, target_ids, target_names)
class IdealPdActuator(Actuator, Generic[IdealPdCfgT]):
"""Ideal PD control actuator."""
param_names = ("stiffness", "damping", "force_limit")
def compute(self, cmd: ActuatorCmd) -> torch.Tensor:
# Shared stateless-law compute. Keeping this (rather than writing a custom
# compute) is what marks an actuator as fusable; subclasses with a different
# control law (e.g. learned networks) override compute instead.
params = {name: getattr(self, name) for name in self.param_names}
return type(self).control_law(params, cmd)
@staticmethod
def control_law(params: dict[str, torch.Tensor], cmd: ActuatorCmd) -> torch.Tensor:
torque = pd_torque(params["stiffness"], params["damping"], cmd)
force_limit = params["force_limit"]
return torch.clamp(torque, -force_limit, force_limit)
def __init__(
self,
cfg: IdealPdCfgT,
entity: Entity,
target_ids: list[int],
target_names: list[str],
) -> None:
super().__init__(cfg, entity, target_ids, target_names)
self.stiffness: torch.Tensor | None = None
self.damping: torch.Tensor | None = None
self.force_limit: torch.Tensor | None = None
self.default_stiffness: torch.Tensor | None = None
self.default_damping: torch.Tensor | None = None
self.default_force_limit: torch.Tensor | None = None
def edit_spec(self, spec: mujoco.MjSpec, target_names: list[str]) -> None:
# Add <motor> actuator to spec, one per target.
for target_name in target_names:
actuator = create_motor_actuator(
spec,
target_name,
effort_limit=self.cfg.effort_limit,
armature=self.cfg.armature,
frictionloss=self.cfg.frictionloss,
viscous_damping=self.cfg.viscous_damping,
transmission_type=self.cfg.transmission_type,
)
self._mjs_actuators.append(actuator)
def initialize(
self,
mj_model: mujoco.MjModel,
model: mjwarp.Model,
data: mjwarp.Data,
device: str,
) -> None:
super().initialize(mj_model, model, data, device)
num_envs = data.nworld
num_joints = len(self._target_names)
self.stiffness = torch.full(
(num_envs, num_joints), self.cfg.stiffness, dtype=torch.float, device=device
)
self.damping = torch.full(
(num_envs, num_joints), self.cfg.damping, dtype=torch.float, device=device
)
self.force_limit = torch.full(
(num_envs, num_joints), self.cfg.effort_limit, dtype=torch.float, device=device
)
self.default_stiffness = self.stiffness.clone()
self.default_damping = self.damping.clone()
self.default_force_limit = self.force_limit.clone()
def set_gains(
self,
env_ids: torch.Tensor | slice,
kp: torch.Tensor | None = None,
kd: torch.Tensor | None = None,
) -> None:
"""Set PD gains for specified environments.
Args:
env_ids: Environment indices to update.
kp: New proportional gains. Shape: (num_envs, num_actuators) or (num_envs,).
kd: New derivative gains. Shape: (num_envs, num_actuators) or (num_envs,).
"""
assert self.stiffness is not None
assert self.damping is not None
if kp is not None:
if kp.ndim == 1:
kp = kp.unsqueeze(-1)
self.stiffness[env_ids] = kp
if kd is not None:
if kd.ndim == 1:
kd = kd.unsqueeze(-1)
self.damping[env_ids] = kd
def set_effort_limit(
self, env_ids: torch.Tensor | slice, effort_limit: torch.Tensor
) -> None:
"""Set effort limits for specified environments.
Args:
env_ids: Environment indices to update.
effort_limit: New effort limits. Shape: (num_envs, num_actuators) or (num_envs,).
"""
assert self.force_limit is not None
if effort_limit.ndim == 1:
effort_limit = effort_limit.unsqueeze(-1)
self.force_limit[env_ids] = effort_limit
+132
View File
@@ -0,0 +1,132 @@
"""Use actuators already defined in your robot XML.
When your MJCF already has ``<position>``, ``<velocity>``, ``<motor>``, or ``<muscle>``
elements, ``XmlActuatorCfg`` wraps them without re-specifying gains or limits. The
actuator type is auto-detected from the spec element. Delay and other ``ActuatorCfg``
fields work as usual.
If your XML has actuators on both joints and tendons, use two configs with different
``transmission_type`` values; a single config can only target one namespace.
"""
from __future__ import annotations
from dataclasses import dataclass
from typing import TYPE_CHECKING, cast
import mujoco
import torch
from mjlab.actuator.actuator import Actuator, ActuatorCfg, ActuatorCmd, CommandField
from mjlab.utils.mujoco import detect_command_field
if TYPE_CHECKING:
from mjlab.entity import Entity
@dataclass(kw_only=True)
class XmlActuatorCfg(ActuatorCfg):
"""Wrap existing XML-defined actuators.
Detects the actuator type (position, velocity, motor, muscle) from the MuJoCo spec
element during ``edit_spec``, or validates against an explicit ``command_field`` if
provided.
"""
command_field: CommandField | None = None
"""If provided, require XML actuators to match this command field.
If None, auto-detect strictly from the spec element."""
def build(
self, entity: Entity, target_ids: list[int], target_names: list[str]
) -> XmlActuator:
return XmlActuator(self, entity, target_ids, target_names)
class XmlActuator(Actuator[XmlActuatorCfg]):
"""Wrapper for XML-defined actuators."""
def __init__(
self,
cfg: XmlActuatorCfg,
entity: Entity,
target_ids: list[int],
target_names: list[str],
) -> None:
super().__init__(cfg, entity, target_ids, target_names)
self._command_field: CommandField | None = None
@property
def command_field(self) -> CommandField:
"""The resolved command field (detected or explicit), not the raw config value.
Set during ``edit_spec``. Use ``self.cfg.command_field`` to access the user's
original input (which may be None for auto-detection).
"""
assert self._command_field is not None
return self._command_field
def edit_spec(self, spec: mujoco.MjSpec, target_names: list[str]) -> None:
# Filter to only targets that have corresponding XML actuators.
filtered_target_ids = []
filtered_target_names = []
for i, target_name in enumerate(target_names):
actuator = self._find_actuator_for_target(spec, target_name)
if actuator is not None:
self._mjs_actuators.append(actuator)
filtered_target_ids.append(self._target_ids_list[i])
# Store the user-facing (stripped) name, not the spec name.
filtered_target_names.append(self._target_names[i])
if len(filtered_target_names) == 0:
raise ValueError(
f"No XML actuators found for any targets matching the patterns. "
f"Searched targets: {target_names}. "
f"XML actuator config expects actuators to already exist in the XML."
)
# Update target IDs and names to only include those with actuators.
self._target_ids_list = filtered_target_ids
self._target_names = filtered_target_names
if self.cfg.command_field is not None:
# Validate that every matched actuator is compatible with the
# requested command field.
for act, name in zip(self._mjs_actuators, self._target_names, strict=True):
try:
detected = detect_command_field(act)
except ValueError:
continue # Unrecognized type; trust the explicit override.
if detected != self.cfg.command_field:
raise ValueError(
f"XML actuator for '{name}' is type '{detected}', but "
f"command_field='{self.cfg.command_field}' was requested."
)
self._command_field = self.cfg.command_field
else:
# Auto-detect from spec elements. Raises for unrecognized types.
fields = [detect_command_field(a) for a in self._mjs_actuators]
unique = set(fields)
if len(unique) > 1:
breakdown = dict(zip(self._target_names, fields, strict=True))
raise ValueError(
f"Mixed XML actuator types for targets: {breakdown}. "
f"Split into separate XmlActuatorCfg groups."
)
self._command_field = cast(CommandField, fields[0])
def compute(self, cmd: ActuatorCmd) -> torch.Tensor:
if self._command_field == "position":
return cmd.position_target
if self._command_field == "velocity":
return cmd.velocity_target
return cmd.effort_target
def _find_actuator_for_target(
self, spec: mujoco.MjSpec, target_name: str
) -> mujoco.MjsActuator | None:
"""Find an actuator that targets the given target."""
for actuator in spec.actuators:
if actuator.target == target_name:
return actuator
return None
+7
View File
@@ -0,0 +1,7 @@
# Asset Zoo
| Robot | Image |
|-------|-------|
| **Unitree G1** | <img src="../../../docs/source/_static/content/g1.png" width="200" height="150" alt="Unitree G1 Humanoid"> |
| **Unitree Go1** | <img src="../../../docs/source/_static/content/go1.png" width="200" height="150" alt="Unitree Go1 Quadruped"> |
| **i2rt YAM** | <img src="../../../docs/source/_static/content/yam.png" width="200" height="150" alt="i2rt YAM Robot Arm"> |
View File
+18
View File
@@ -0,0 +1,18 @@
from mjlab.asset_zoo.robots.i2rt_yam.yam_constants import (
YAM_ACTION_SCALE as YAM_ACTION_SCALE,
)
from mjlab.asset_zoo.robots.i2rt_yam.yam_constants import (
get_yam_robot_cfg as get_yam_robot_cfg,
)
from mjlab.asset_zoo.robots.unitree_g1.g1_constants import (
G1_ACTION_SCALE as G1_ACTION_SCALE,
)
from mjlab.asset_zoo.robots.unitree_g1.g1_constants import (
get_g1_robot_cfg as get_g1_robot_cfg,
)
from mjlab.asset_zoo.robots.unitree_go1.go1_constants import (
GO1_ACTION_SCALE as GO1_ACTION_SCALE,
)
from mjlab.asset_zoo.robots.unitree_go1.go1_constants import (
get_go1_robot_cfg as get_go1_robot_cfg,
)
@@ -0,0 +1 @@
"""Booster T1 external asset adapter."""
@@ -0,0 +1,28 @@
BSD 3-Clause License
Copyright (c) 2025, BoosterRobotics
Redistribution and use in source and binary forms, with or without
modification, are permitted provided that the following conditions are met:
1. Redistributions of source code must retain the above copyright notice, this
list of conditions and the following disclaimer.
2. Redistributions in binary form must reproduce the above copyright notice,
this list of conditions and the following disclaimer in the documentation
and/or other materials provided with the distribution.
3. Neither the name of the copyright holder nor the names of its
contributors may be used to endorse or promote products derived from
this software without specific prior written permission.
THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE
FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
@@ -0,0 +1,216 @@
<mujoco model="T1">
<compiler angle="radian" meshdir="meshes/" autolimits="true"/>
<asset>
<texture type="skybox" builtin="gradient" rgb1="0.3 0.5 0.7" rgb2="0 0 0" width="512" height="512"/>
<texture name="texplane" type="2d" builtin="checker" rgb1=".2 .3 .4" rgb2=".1 0.15 0.2" width="512" height="512" mark="cross" markrgb=".8 .8 .8"/>
<material name="matplane" reflectance="0.3" texture="texplane" texrepeat="1 1" texuniform="true"/>
<mesh name="Trunk" file="Trunk.STL"/>
<mesh name="H1" file="H1.STL"/>
<mesh name="H2" file="H2.STL"/>
<mesh name="AL1" file="AL1.STL"/>
<mesh name="AL2" file="AL2.STL"/>
<mesh name="AL3" file="AL3.STL"/>
<mesh name="left_hand_link" file="left_hand_link.STL"/>
<mesh name="AR1" file="AR1.STL"/>
<mesh name="AR2" file="AR2.STL"/>
<mesh name="AR3" file="AR3.STL"/>
<mesh name="right_hand_link" file="right_hand_link.STL"/>
<mesh name="Waist" file="Waist.STL"/>
<mesh name="Hip_Pitch_Left" file="Hip_Pitch_Left.STL"/>
<mesh name="Hip_Roll_Left" file="Hip_Roll_Left.STL"/>
<mesh name="Hip_Yaw_Left" file="Hip_Yaw_Left.STL"/>
<mesh name="Shank_Left" file="Shank_Left.STL"/>
<mesh name="Ankle_Cross_Left" file="Ankle_Cross_Left.STL"/>
<mesh name="left_foot_link" file="left_foot_link.STL"/>
<mesh name="Hip_Pitch_Right" file="Hip_Pitch_Right.STL"/>
<mesh name="Hip_Roll_Right" file="Hip_Roll_Right.STL"/>
<mesh name="Hip_Yaw_Right" file="Hip_Yaw_Right.STL"/>
<mesh name="Shank_Right" file="Shank_Right.STL"/>
<mesh name="Ankle_Cross_Right" file="Ankle_Cross_Right.STL"/>
<mesh name="right_foot_link" file="right_foot_link.STL"/>
<mesh name="Logo" file="Logo.STL"/>
</asset>
<worldbody>
<light directional="true" diffuse=".4 .4 .4" specular="0.1 0.1 0.1" pos="0 0 5.0" dir="0 0 -1" castshadow="false" />
<light directional="true" diffuse=".6 .6 .6" specular="0.2 0.2 0.2" pos="0 0 4" dir="0 0 -1" />
<geom name="ground" type="plane" pos="0 0 0" size="0 0 1" material="matplane" condim="1"/>
<body name="trunk" pos="0 0 0.7">
<inertial pos="0.055138 -1.42184e-06 0.105043" quat="0.999959 1.39867e-05 -0.00902028 -1.39205e-05" mass="11.7016" diaginertia="0.0915734 0.0768118 0.0556067"/>
<site name='imu' size='0.01' pos='0.0 0.0 0.0' />
<joint name="world_joint" type="free" limited="false" actuatorfrclimited="false"/>
<geom type="mesh" contype="0" conaffinity="0" group="1" density="0" rgba="0.76 0.76 0.76 1." mesh="Trunk"/>
<geom size="0.085 0.09 0.145" pos="0.06 0 0.12" type="box" rgba="0.76 0.76 0.76 0."/>
<geom type="mesh" contype="0" conaffinity="0" group="1" density="0" rgba="0.1 0.1 0.1 1" mesh="Logo"/>
<body name="aahead_yaw_link" pos="0.0625 0 0.243">
<inertial pos="-0.000508 -0.001403 0.057432" quat="0.763881 -0.0132172 0.00173419 0.645219" mass="0.44391" diaginertia="0.000241549 0.00022351 0.000149941"/>
<joint name="aahead_yaw_joint" pos="0 0 0" axis="0 0 1" range="-1.309 1.309" armature="0.0018"/>
<geom type="mesh" contype="0" conaffinity="0" group="1" density="0" rgba="0.4 0.4 0.4 1." mesh="H1"/>
<geom size="0.025 0.035" pos="0 0 0.06" quat="0.707105 0.707108 0 0" type="cylinder" rgba="0.4 0.4 0.4 0."/>
<body name="aahead_pitch_link" pos="0 0 0.06185">
<inertial pos="0.007802 0.001262 0.098631" quat="0.988453 0.106172 -0.0745686 -0.0782786" mass="0.631019" diaginertia="0.00203553 0.00192467 0.00172381"/>
<joint name="aahead_pitch_joint" pos="0 0 0" axis="0 1 0" range="-0.35 0.96" armature="0.0018"/>
<geom type="mesh" contype="0" conaffinity="0" group="1" density="0" rgba="0.4 0.4 0.4 1." mesh="H2"/>
<geom size="0.08" pos="0.01 0 0.11" rgba="0.4 0.4 0.4 0.5"/>
<body name="head_realsense_rgb_link" pos="0.063961 -0.0115 0.108406" quat="0.46390386 -0.53366035 0.53366018 -0.46390371"/>
</body>
</body>
<body name="left_shoulder_pitch_link" pos="0.0575 0.1063 0.219" quat="1 0 0.000440565 0">
<inertial pos="-0.000677 0.044974 0" quat="0.50423 0.495734 -0.50423 0.495734" mass="0.53" diaginertia="0.001367 0.00129329 0.000292711"/>
<joint name="left_shoulder_pitch_joint" pos="0 0 0" axis="0 1 0" range="-3.276 1.196" armature="0.0282528"/>
<geom type="mesh" contype="0" conaffinity="0" group="1" density="0" rgba="0.76 0.76 0.76 1." mesh="AL1"/>
<geom size="0.03 0.035" pos="0 0.047 0" quat="0.707105 0 0.707108 0" type="cylinder" rgba="0.76 0.76 0.76 0."/>
<body name="left_shoulder_roll_link" pos="0 0.047 0">
<inertial pos="0.003862 0.037976 0" quat="0.487991 0.511727 -0.487991 0.511727" mass="0.16" diaginertia="0.000401 0.00034538 0.00017662"/>
<joint name="left_shoulder_roll_joint" pos="0 0 0" axis="1 0 0" range="-1.623 1.736" armature="0.0282528"/>
<geom type="mesh" contype="0" conaffinity="0" group="1" density="0" rgba="0.4 0.4 0.4 1." mesh="AL2"/>
<body name="left_elbow_pitch_link" pos="0.00025 0.0605 0">
<inertial pos="0 0.085353 -9.9e-05" quat="0.70641 0.707803 0 0" mass="1.02" diaginertia="0.012869 0.012798 0.000620953"/>
<joint name="left_elbow_pitch_joint" pos="0 0 0" axis="0 1 0" range="-2.235 2.235" armature="0.0282528"/>
<geom type="mesh" contype="0" conaffinity="0" group="1" density="0" rgba="0.4 0.4 0.4 1." mesh="AL3"/>
<geom size="0.03 0.095" pos="0 0.07 0" quat="0.707105 0.707108 0 0" type="cylinder" rgba="0.4 0.4 0.4 0."/>
<body name="left_elbow_yaw_link" pos="0 0.1471 0">
<inertial pos="-0.000107344 0.11022 0.00058741" quat="0.707869 0.706344 -5.3547e-05 -0.000214933" mass="0.329214" diaginertia="0.00819152 0.00816355 0.000224963"/>
<joint name="left_elbow_yaw_joint" pos="0 0 0" axis="0 0 1" range="-2.077 1.658" armature="0.0282528"/>
<geom type="mesh" contype="0" conaffinity="0" group="1" density="0" rgba="0.4 0.4 0.4 1." mesh="left_hand_link"/>
<geom size="0.029 0.065" pos="0 0.12 0" quat="0.707105 0.707108 0 0" type="cylinder" rgba="0.4 0.4 0.4 0."/>
<geom size="0.031" pos="0 0.216 0"/>
</body>
</body>
</body>
</body>
<body name="right_shoulder_pitch_link" pos="0.0575 -0.1063 0.219" quat="1 0 0.000440565 0">
<inertial pos="-0.000677 -0.044974 0" quat="0.50423 0.495734 -0.50423 0.495734" mass="0.53" diaginertia="0.001367 0.00129329 0.000292711"/>
<joint name="right_shoulder_pitch_joint" pos="0 0 0" axis="0 1 0" range="-3.276 1.196" armature="0.0282528"/>
<geom type="mesh" contype="0" conaffinity="0" group="1" density="0" rgba="0.76 0.76 0.76 1." mesh="AR1"/>
<geom size="0.03 0.035" pos="0 -0.047 0" quat="0.707105 0 0.707108 0" type="cylinder" rgba="0.76 0.76 0.76 0."/>
<body name="right_shoulder_roll_link" pos="0 -0.047 0">
<inertial pos="0.003862 -0.037976 0" quat="0.511727 0.487991 -0.511727 0.487991" mass="0.16" diaginertia="0.000401 0.00034538 0.00017662"/>
<joint name="right_shoulder_roll_joint" pos="0 0 0" axis="1 0 0" range="-1.736 1.623" armature="0.0282528"/>
<geom type="mesh" contype="0" conaffinity="0" group="1" density="0" rgba="0.4 0.4 0.4 1." mesh="AR2"/>
<body name="right_elbow_pitch_link" pos="0.00025 -0.0605 0">
<inertial pos="0 -0.085353 -9.9e-05" quat="0.707803 0.70641 0 0" mass="1.02" diaginertia="0.012869 0.012798 0.000620953"/>
<joint name="right_elbow_pitch_joint" pos="0 0 0" axis="0 1 0" range="-2.235 2.235" armature="0.0282528"/>
<geom type="mesh" contype="0" conaffinity="0" group="1" density="0" rgba="0.4 0.4 0.4 1." mesh="AR3"/>
<geom size="0.03 0.095" pos="0 -0.07 0" quat="0.707105 0.707108 0 0" type="cylinder" rgba="0.4 0.4 0.4 0."/>
<body name="right_elbow_yaw_link" pos="0 -0.1471 0">
<inertial pos="-0.000107344 -0.11022 0.00058741" quat="0.706344 0.707869 0.000214933 5.3547e-05" mass="0.329214" diaginertia="0.00819152 0.00816355 0.000224963"/>
<joint name="right_elbow_yaw_joint" pos="0 0 0" axis="0 0 1" range="-1.658 2.077" armature="0.0282528"/>
<geom type="mesh" contype="0" conaffinity="0" group="1" density="0" rgba="0.4 0.4 0.4 1." mesh="right_hand_link"/>
<geom size="0.029 0.065" pos="0 -0.12 0" quat="0.707105 0.707108 0 0" type="cylinder" rgba="0.4 0.4 0.4 0."/>
<geom size="0.031" pos="0 -0.216 0"/>
</body>
</body>
</body>
</body>
<body name="waist_yaw_link" pos="0.0625 0 -0.1155">
<inertial pos="0.002284 3e-06 0.007301" quat="0.983649 0.000360386 -0.180076 0.00269791" mass="2.581" diaginertia="0.00536742 0.005299 0.00474258"/>
<joint name="waist_yaw_joint" pos="0 0 0" axis="0 0 1" range="-1.012 1.012" armature="0.0478125"/>
<geom type="mesh" contype="0" conaffinity="0" group="1" density="0" rgba="0.4 0.4 0.4 1." mesh="Waist"/>
<geom size="0.06 0.05" quat="0.707105 0.707108 0 0" type="cylinder" rgba="0.4 0.4 0.4 0."/>
<body name="left_hip_pitch_link" pos="0 0.106 0">
<inertial pos="0.000534 -0.007296 -0.018083" quat="0.975141 0.2211 0.0145808 0.0017406" mass="1.021" diaginertia="0.00180547 0.00145926 0.00125327"/>
<joint name="left_hip_pitch_joint" pos="0 0 0" axis="0 1 0" range="-2.967 2.191" armature="0.0523908"/>
<geom type="mesh" contype="0" conaffinity="0" group="1" density="0" rgba="0.76 0.76 0.76 1." mesh="Hip_Pitch_Left"/>
<geom size="0.045 0.0425" pos="-0.002 0 -0.02" quat="0.707105 0 0.707108 0" type="cylinder" rgba="0.76 0.76 0.76 0."/>
<body name="left_hip_roll_link" pos="0 0 -0.02">
<inertial pos="0.001101 2.4e-05 -0.05375" quat="0.707081 -0.00599604 -0.00599604 0.707081" mass="0.385" diaginertia="0.001743 0.00151729 0.000514712"/>
<joint name="left_hip_roll_joint" pos="0 0 0" axis="1 0 0" range="-0.363 1.536" armature="0.0478125"/>
<geom type="mesh" contype="0" conaffinity="0" group="1" density="0" rgba="0.4 0.4 0.4 1." mesh="Hip_Roll_Left"/>
<geom size="0.045 0.04" pos="0 0 -0.11" type="cylinder" rgba="0.4 0.4 0.4 0."/>
<body name="left_hip_yaw_link" pos="0 0 -0.081854">
<inertial pos="-0.007233 0.000206 -0.089184" quat="0.696808 -0.033953 -0.0315708 0.715758" mass="2.166" diaginertia="0.0257334 0.0253024 0.00259215"/>
<joint name="left_hip_yaw_joint" pos="0 0 0" axis="0 0 1" range="-1.076 1.062" armature="0.0478125"/>
<geom type="mesh" contype="0" conaffinity="0" group="1" density="0" rgba="0.4 0.4 0.4 1." mesh="Hip_Yaw_Left"/>
<geom size="0.055 0.045" pos="-0.015 0 -0.135" quat="0.707105 0.707108 0 0" type="cylinder" rgba="0.4 0.4 0.4 0."/>
<body name="left_knee_pitch_link" pos="-0.014 0 -0.134">
<inertial pos="-0.006012 0.000259 -0.124318" quat="0.997573 0.00144024 -0.023949 0.0653706" mass="1.82" diaginertia="0.0346951 0.0345375 0.00185844"/>
<joint name="left_knee_pitch_joint" pos="0 0 0" axis="0 1 0" range="0 2.145" armature="0.0636012"/>
<geom type="mesh" contype="0" conaffinity="0" group="1" density="0" rgba="0.4 0.4 0.4 1." mesh="Shank_Left"/>
<geom size="0.05 0.04 0.07" pos="0 0 -0.15" type="box" rgba="0.4 0.4 0.4 0."/>
<body name="left_ankle_pitch_link" pos="0 0 -0.28">
<inertial pos="-0.003722 0 -0.007981" quat="0.443136 0.551027 0.551027 0.443136" mass="0.073" diaginertia="2.9e-05 2.56589e-05 1.13411e-05"/>
<joint name="left_ankle_pitch_joint" pos="0 0 0" axis="0 1 0" range="-0.87 0.35" armature="0.0679104"/>
<geom type="mesh" contype="0" conaffinity="0" group="1" density="0" rgba="0.4 0.4 0.4 1." mesh="Ankle_Cross_Left"/>
<body name="left_ankle_roll_link" pos="0 0.00025 -0.012">
<inertial pos="-0.000249 0 -0.00914" quat="0 0.651138 0 0.75896" mass="0.685" diaginertia="0.00269361 0.002385 0.00171539"/>
<joint name="left_ankle_roll_joint" pos="0 0 0" axis="1 0 0" range="-0.44 0.44" armature="0.02037312"/>
<geom type="mesh" contype="0" conaffinity="0" group="1" density="0" rgba="0.4 0.4 0.4 1." mesh="left_foot_link"/>
<geom size="0.112434 0.05 0.02183" pos="0.0101079 0 -0.0214208" type="box" rgba="0.4 0.4 0.4 0."/>
</body>
</body>
</body>
</body>
</body>
</body>
<body name="right_hip_pitch_link" pos="0 -0.106 0">
<inertial pos="0.000534 0.007514 -0.018082" quat="0.973446 -0.228403 0.0147148 -0.00419349" mass="1.021" diaginertia="0.00180552 0.0014632 0.00124928"/>
<joint name="right_hip_pitch_joint" pos="0 0 0" axis="0 1 0" range="-2.967 2.191" armature="0.0523908"/>
<geom type="mesh" contype="0" conaffinity="0" group="1" density="0" rgba="0.76 0.76 0.76 1." mesh="Hip_Pitch_Right"/>
<geom size="0.045 0.0425" pos="-0.002 0 -0.02" quat="0.707105 0 0.707108 0" type="cylinder" rgba="0.76 0.76 0.76 0."/>
<body name="right_hip_roll_link" pos="0 0 -0.02">
<inertial pos="0.001099 2.4e-05 -0.053748" quat="0.707081 -0.00599604 -0.00599604 0.707081" mass="0.385" diaginertia="0.001743 0.00151729 0.000514712"/>
<joint name="right_hip_roll_joint" pos="0 0 0" axis="1 0 0" range="-1.536 0.363" armature="0.0478125"/>
<geom type="mesh" contype="0" conaffinity="0" group="1" density="0" rgba="0.4 0.4 0.4 1." mesh="Hip_Roll_Right"/>
<geom size="0.045 0.04" pos="0 0 -0.11" type="cylinder" rgba="0.4 0.4 0.4 0."/>
<body name="right_hip_yaw_link" pos="0 0 -0.081854">
<inertial pos="-0.007191 -0.000149 -0.08922" quat="0.714468 -0.0315638 -0.0336391 0.698146" mass="2.17" diaginertia="0.0257623 0.0253298 0.00259389"/>
<joint name="right_hip_yaw_joint" pos="0 0 0" axis="0 0 1" range="-1.076 1.062" armature="0.0478125"/>
<geom type="mesh" contype="0" conaffinity="0" group="1" density="0" rgba="0.4 0.4 0.4 1." mesh="Hip_Yaw_Right"/>
<geom size="0.055 0.045" pos="-0.015 0 -0.135" quat="0.707105 0.707108 0 0" type="cylinder" rgba="0.4 0.4 0.4 0."/>
<body name="right_knee_pitch_link" pos="-0.014 0 -0.134">
<inertial pos="-0.005741 -0.000541 -0.122602" quat="0.99926 -0.000580963 -0.023461 -0.0304754" mass="1.82" diaginertia="0.0351717 0.0349574 0.00196589"/>
<joint name="right_knee_pitch_joint" pos="0 0 0" axis="0 1 0" range="0 2.145" armature="0.0636012"/>
<geom type="mesh" contype="0" conaffinity="0" group="1" density="0" rgba="0.4 0.4 0.4 1." mesh="Shank_Right"/>
<geom size="0.05 0.04 0.07" pos="0 0 -0.15" type="box" rgba="0.4 0.4 0.4 0."/>
<body name="right_ankle_pitch_link" pos="0 0 -0.28">
<inertial pos="-0.003722 0 -0.007981" quat="0.443136 0.551027 0.551027 0.443136" mass="0.073" diaginertia="2.9e-05 2.56589e-05 1.13411e-05"/>
<joint name="right_ankle_pitch_joint" pos="0 0 0" axis="0 1 0" range="-0.87 0.35" armature="0.0679104"/>
<geom type="mesh" contype="0" conaffinity="0" group="1" density="0" rgba="0.4 0.4 0.4 1." mesh="Ankle_Cross_Right"/>
<body name="right_ankle_roll_link" pos="0 -0.00025 -0.012">
<inertial pos="-0.000248 0 -0.00914" quat="0 0.651138 0 0.75896" mass="0.685" diaginertia="0.00269361 0.002385 0.00171539"/>
<joint name="right_ankle_roll_joint" pos="0 0 0" axis="1 0 0" range="-0.44 0.44" armature="0.02037312"/>
<geom type="mesh" contype="0" conaffinity="0" group="1" density="0" rgba="0.4 0.4 0.4 1." mesh="right_foot_link"/>
<geom size="0.112434 0.05 0.02183" pos="0.0101079 0 -0.0214208" type="box" rgba="0.4 0.4 0.4 0."/>
</body>
</body>
</body>
</body>
</body>
</body>
</body>
</body>
</worldbody>
<actuator>
<motor name="aahead_yaw_joint" joint="aahead_yaw_joint" forcerange="-7.0 7.0"/>
<motor name="aahead_pitch_joint" joint="aahead_pitch_joint" forcerange="-7.0 7.0"/>
<motor name="left_shoulder_pitch_joint" joint="left_shoulder_pitch_joint" forcerange="-38.3 38.3"/>
<motor name="left_shoulder_roll_joint" joint="left_shoulder_roll_joint" forcerange="-38.3 38.3"/>
<motor name="left_elbow_pitch_joint" joint="left_elbow_pitch_joint" forcerange="-38.3 38.3"/>
<motor name="left_elbow_yaw_joint" joint="left_elbow_yaw_joint" forcerange="-38.3 38.3"/>
<motor name="right_shoulder_pitch_joint" joint="right_shoulder_pitch_joint" forcerange="-38.3 38.3"/>
<motor name="right_shoulder_roll_joint" joint="right_shoulder_roll_joint" forcerange="-38.3 38.3"/>
<motor name="right_elbow_pitch_joint" joint="right_elbow_pitch_joint" forcerange="-38.3 38.3"/>
<motor name="right_elbow_yaw_joint" joint="right_elbow_yaw_joint" forcerange="-38.3 38.3"/>
<motor name="waist_yaw_joint" joint="waist_yaw_joint" forcerange="-68.0 68.0"/>
<motor name="left_hip_pitch_joint" joint="left_hip_pitch_joint" forcerange="-98.8 98.8"/>
<motor name="left_hip_roll_joint" joint="left_hip_roll_joint" forcerange="-68.0 68.0"/>
<motor name="left_hip_yaw_joint" joint="left_hip_yaw_joint" forcerange="-68.0 68.0"/>
<motor name="left_knee_pitch_joint" joint="left_knee_pitch_joint" forcerange="-130.5 130.5"/>
<motor name="left_ankle_pitch_joint" joint="left_ankle_pitch_joint" forcerange="-73.1 73.1"/>
<motor name="left_ankle_roll_joint" joint="left_ankle_roll_joint" forcerange="-17.2 17.2"/>
<motor name="right_hip_pitch_joint" joint="right_hip_pitch_joint" forcerange="-98.8 98.8"/>
<motor name="right_hip_roll_joint" joint="right_hip_roll_joint" forcerange="-68.0 68.0"/>
<motor name="right_hip_yaw_joint" joint="right_hip_yaw_joint" forcerange="-68.0 68.0"/>
<motor name="right_knee_pitch_joint" joint="right_knee_pitch_joint" forcerange="-130.5 130.5"/>
<motor name="right_ankle_pitch_joint" joint="right_ankle_pitch_joint" forcerange="-73.1 73.1"/>
<motor name="right_ankle_roll_joint" joint="right_ankle_roll_joint" forcerange="-17.2 17.2"/>
</actuator>
<sensor>
<framequat name="orientation" objtype="site" noise="0.001" objname="imu" />
<gyro name="angular-velocity" site="imu" noise="0.005" />
</sensor>
</mujoco>
@@ -0,0 +1,144 @@
"""Booster T1 (23 DoF), using bundled BSD-licensed assets by default.
[uh-vendor] External BOOSTER_ASSETS_ROOT is an explicit optional model override.
Inertias, limits, armatures and torque ratings come from T1_23dof.xml.
PD gains are simulation training defaults, not a hardware deployment recipe.
"""
import os
from pathlib import Path
import mujoco
from mjlab.actuator import BuiltinPositionActuatorCfg
from mjlab.entity import EntityArticulationInfoCfg, EntityCfg
from mjlab.utils.spec_config import CollisionCfg
def get_t1_xml() -> Path:
root = os.environ.get("BOOSTER_ASSETS_ROOT")
if root:
return Path(root).expanduser() / "robots/T1/T1_23dof.xml"
return Path(__file__).parent / "assets/T1_23dof.xml"
def get_spec() -> mujoco.MjSpec:
path = get_t1_xml()
if not path.is_file():
raise FileNotFoundError(
f"T1 asset not found: {path}. Check the bundled assets or BOOSTER_ASSETS_ROOT override."
)
spec = mujoco.MjSpec.from_file(str(path))
# The scene owns the ground. Replace torque motors, rather than duplicating them.
spec.delete(spec.geom("ground"))
for actuator in list(spec.actuators):
spec.delete(actuator)
for sensor in list(spec.sensors):
spec.delete(sensor)
for light in list(spec.lights):
spec.delete(light)
for body in spec.bodies:
for i, geom in enumerate(body.geoms):
if geom.type == mujoco.mjtGeom.mjGEOM_MESH:
geom.name = f"{body.name}_visual_{i}"
else:
geom.name = f"{body.name}_collision_{i}"
geom.group = 3
for side in ("left", "right"):
foot = spec.body(f"{side}_ankle_roll_link")
foot.add_site(
name=f"{side}_foot",
pos=(0.0101079, 0.0, -0.0432508),
size=(0.01, 0.01, 0.01),
)
for name, sensor_type in (
("imu_lin_vel", mujoco.mjtSensor.mjSENS_VELOCIMETER),
("imu_ang_vel", mujoco.mjtSensor.mjSENS_GYRO),
):
spec.add_sensor(
name=name, type=sensor_type, objtype=mujoco.mjtObj.mjOBJ_SITE, objname="imu"
)
spec.add_sensor(
name="root_angmom",
type=mujoco.mjtSensor.mjSENS_SUBTREEANGMOM,
objtype=mujoco.mjtObj.mjOBJ_BODY,
objname="trunk",
)
return spec
# XML armatures are preserved (no armature override).
# Each tuple is (joint expression, kp, kd, XML torque limit).
_MOTORS = (
("aahead_.*_joint", 5.0, 0.5, 7.0),
(".*_shoulder_.*_joint", 40.0, 2.0, 38.3),
(".*_elbow_.*_joint", 40.0, 2.0, 38.3),
("waist_yaw_joint", 100.0, 3.0, 68.0),
(".*_hip_pitch_joint", 200.0, 5.0, 98.8),
(".*_hip_roll_joint", 150.0, 4.0, 68.0),
(".*_hip_yaw_joint", 100.0, 3.0, 68.0),
(".*_knee_pitch_joint", 200.0, 5.0, 130.5),
(".*_ankle_pitch_joint", 50.0, 2.0, 73.1),
(".*_ankle_roll_joint", 20.0, 1.0, 17.2),
)
T1_ACTION_SCALE = {pattern: 0.25 * effort / kp for pattern, kp, _, effort in _MOTORS}
T1_FOOT_GEOMS = tuple(
f"{side}_ankle_roll_link_collision_1" for side in ("left", "right")
)
def get_t1_robot_cfg() -> EntityCfg:
"""Return independent train/play configs; defer external asset IO until build."""
return EntityCfg(
spec_fn=get_spec,
sort_actuators=True,
init_state=EntityCfg.InitialStateCfg(
pos=(0.0, 0.0, 0.6811),
joint_pos={
".*_hip_pitch_joint": -0.2,
".*_knee_pitch_joint": 0.4,
".*_ankle_pitch_joint": -0.2,
"left_shoulder_roll_joint": -1.3,
"right_shoulder_roll_joint": 1.3,
".*_shoulder_pitch_joint": 0.2,
"left_elbow_yaw_joint": -0.4,
"right_elbow_yaw_joint": 0.4,
},
joint_vel={".*": 0.0},
),
articulation=EntityArticulationInfoCfg(
actuators=tuple(
BuiltinPositionActuatorCfg(
target_names_expr=(pattern,),
stiffness=kp,
damping=kd,
effort_limit=effort,
)
for pattern, kp, kd, effort in _MOTORS
),
soft_joint_pos_limit_factor=0.9,
),
# Terrain contacts everywhere; opposite legs also collide. Same-chain CAD
# overlaps and arm/torso self contacts are deliberately excluded.
collisions=(
CollisionCfg(
geom_names_expr=(".*_collision_.*",),
contype={
"left_(hip|knee|ankle).*_collision_.*": 2,
"right_(hip|knee|ankle).*_collision_.*": 4,
".*": 0,
},
conaffinity={
"left_(hip|knee|ankle).*_collision_.*": 5,
"right_(hip|knee|ankle).*_collision_.*": 3,
".*": 1,
},
condim={".*_ankle_roll_link_collision_.*": 3, ".*": 1},
priority={".*_ankle_roll_link_collision_.*": 1, ".*": 0},
friction=(0.8,),
),
),
)

Some files were not shown because too many files have changed in this diff Show More