Files
Mujoco_WASM/mjx/mujoco/mjx/test_data/humanoid/README.md
T
Erik Frey e59605135a Make benchmarks and viewer more accessible.
- Drop google_benchmark module dependency, wasn't really using it.
- Move benchmark models into test_data
- Rename benchmark => testspeed, make default output more like MJ testspeed.
- Update viewer to use new io functions and add get_data_into.
- Add "mjx-testspeed" and "mjx-viewer" bin scripts

PiperOrigin-RevId: 601277135
Change-Id: I782f666c58a96ac1a0a1c5ffc8ecf29d9fcd810f
2024-01-24 17:04:02 -08:00

909 B

Humanoid

Degrees of Freedom: 27 Actuators: 21

This is a clone of the MuJoCo Humanoid with the following changes:

  • Solver switched to CG with 8 iterations
  • Explicit contact pairs for feet and ground (compatible with OpenAI Gym Humanoid environment)

This simplified humanoid model, introduced in [1], is designed for bipedal locomotion behaviours. While several variants of it exist in the wild, this version is based on the model in the DeepMind Control Suite [2], which has fairly realistic actuator gains.

[1] [Synthesis and Stabilization of Complex Behaviors through Online Trajectory Optimization] (https://doi.org/10.1109/IROS.2012.6386025).

[2] DeepMind Control Suite.

humanoid