Add collisions between flexes and SDF geoms.

The flex-SDF algorithm uses the same logic as mesh-SDF. Previous to this change, the flex was instead colliding with the convex hull of the mesh generated automatically with marching cubes from the SDF.

PiperOrigin-RevId: 876420302
Change-Id: I3bf6552b6b8e108ba4c4209bbcf10fe5f2b2fd90

doc/changelog.rst

@@ -5,6 +5,11 @@ Changelog
 Upcoming version (not yet released)
 -----------------------------------
 
+General
+^^^^^^^
+
+- Flexes now support collisions with SDF geoms.
+
 MJX
 ^^^
 

model/flex/cloth_sdf.xml

@@ -0,0 +1,51 @@
+<!-- Copyright 2026 DeepMind Technologies Limited
+
+     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.
+-->
+
+<mujoco>
+  <extension>
+    <plugin plugin="mujoco.sdf.torus">
+      <instance name="torus">
+        <config key="radius1" value="0.55"/>
+        <config key="radius2" value="0.15"/>
+      </instance>
+    </plugin>
+  </extension>
+
+  <asset>
+    <mesh name="torus">
+      <plugin instance="torus"/>
+    </mesh>
+  </asset>
+
+  <include file="scene.xml"/>
+
+  <worldbody>
+    <body pos="0 0 0.3">
+      <geom type="sdf" mesh="torus" rgba=".8 .17 .15 1">
+        <plugin instance="torus"/>
+      </geom>
+    </body>
+    <body name="flex" pos="0 0 .6">
+      <flexcomp name="test" type="grid" count="8 8 1" spacing=".2 .2 .2" dim="2" rgba="0 .7 .7 1">
+        <edge equality="true"/>
+        <contact selfcollide="none" internal="false"/>
+      </flexcomp>
+    </body>
+    <body pos="0 0 1.0">
+      <freejoint/>
+      <geom type="sphere" size="0.1" mass=".5" rgba="0 0 1 .2"/>
+    </body>
+  </worldbody>
+</mujoco>

src/engine/engine_collision_driver.c

@@ -484,6 +484,14 @@ void mj_collision(const mjModel* m, mjData* d) {
             continue;
           }
 
+          // SDF special processing
+          if (m->geom_type[g] == mjGEOM_SDF) {
+            int ncon_before = d->ncon;
+            mj_collideSdfFlex(m, d, g, f);
+            filterFlexContacts(d, ncon_before);
+            continue;
+          }
+
           // collide geom with flex elements
           int ncon_before = d->ncon;
           int elemnum = m->flex_elemnum[f];
@@ -754,6 +762,15 @@ void mj_collideTree(const mjModel* m, mjData* d, int bf1, int bf2,
     }
   }
 
+  // for body:flex, if body has SDFs, call mj_collideSdfFlex directly
+  if (isbody1 && !isbody2) {
+    for (int i=m->body_geomadr[bf1]; i < m->body_geomadr[bf1]+m->body_geomnum[bf1]; i++) {
+      if (m->geom_type[i] == mjGEOM_SDF) {
+        mj_collideSdfFlex(m, d, i, f2);
+      }
+    }
+  }
+
   // collide trees
   while (nstack) {
     // pop from stack
@@ -820,8 +837,9 @@ void mj_collideTree(const mjModel* m, mjData* d, int bf1, int bf2,
                             d->geom_xpos + 3*nodeid1, d->geom_xmat + 9*nodeid1,
                             NULL, NULL,
                             margin, NULL, NULL, &initialize)) {
-            // collide unless geom is plane (plane:flex handled separately)
+            // collide unless geom is plane or SDF (handled separately)
-            if (m->geom_type[nodeid1] != mjGEOM_PLANE) {
+            if (m->geom_type[nodeid1] != mjGEOM_PLANE &&
+                m->geom_type[nodeid1] != mjGEOM_SDF) {
               mj_collideGeomElem(m, d, nodeid1, f2, nodeid2);
             }
             if (mark_active) {
@@ -1800,6 +1818,54 @@ void mj_collidePlaneFlex(const mjModel* m, mjData* d, int g, int f) {
 }
 
 
+// test an SDF geom and a flex for collision, add to contact list
+void mj_collideSdfFlex(const mjModel* m, mjData* d, int g, int f) {
+  // only support dim==2 (triangular elements)
+  if (m->flex_dim[f] != 2) {
+    return;
+  }
+
+  // prepare contact parameters (same for all contacts)
+  mjtNum margin = mj_assignMargin(m, m->geom_margin[g] + m->flex_margin[f]);
+  int condim;
+  mjtNum gap, solref[mjNREF], solimp[mjNIMP], friction[5];
+  mjtNum solreffriction[mjNREF] = {0};
+  mj_contactParam(m, &condim, &gap, solref, solimp, friction, g, -1, -1, f);
+
+  // allocate temporary contact array on stack (zero-initialized)
+  mj_markStack(d);
+  mjContact* con = mjSTACKALLOC(d, mjMAXCONPAIR, mjContact);
+  memset(con, 0, mjMAXCONPAIR * sizeof(mjContact));
+
+  // call batched flex-SDF collision
+  int num = mjc_FlexSDF(m, d, con, g, f, margin);
+
+  // add contacts to mjData
+  for (int i = 0; i < num; i++) {
+    // set contact ids
+    con[i].geom[0] = g;
+    con[i].geom[1] = -1;
+    con[i].flex[0] = -1;
+    con[i].flex[1] = f;
+    con[i].elem[0] = -1;
+    // con[i].elem[1] already set by mjc_FlexSDF with actual element index
+    con[i].vert[0] = -1;
+    con[i].vert[1] = -1;
+
+    // set remaining contact parameters
+    mj_setContact(m, con + i, condim, margin-gap, solref, solreffriction, solimp, friction);
+
+    // add to mjData, abort if too many contacts
+    if (mj_addContact(m, d, con + i)) {
+      mj_freeStack(d);
+      return;
+    }
+  }
+
+  mj_freeStack(d);
+}
+
+
 // test single triangle plane : vertex
 static int planeVertex(mjContact* con, const mjtNum* pos, mjtNum rad,
                        int t0, int t1, int t2, int v) {

src/engine/engine_collision_driver.h

@@ -55,6 +55,9 @@ void mj_collideGeoms(const mjModel* m, mjData* d, int g1, int g2);
 // test a plane geom and a flex for collision, add to contact list
 void mj_collidePlaneFlex(const mjModel* m, mjData* d, int g, int f);
 
+// test an SDF geom and a flex for collision, add to contact list
+void mj_collideSdfFlex(const mjModel* m, mjData* d, int g, int f);
+
 // test for internal flex collisions, add to contact list
 void mj_collideFlexInternal(const mjModel* m, mjData* d, int f);
 

src/engine/engine_collision_sdf.c

@@ -22,9 +22,7 @@
 #include <mujoco/mjsan.h>  // IWYU pragma: keep
 #include <mujoco/mjtnum.h>
 #include "engine/engine_collision_primitive.h"
-#include "engine/engine_memory.h"
 #include "engine/engine_plugin.h"
-#include "engine/engine_sort.h"
 #include "engine/engine_util_blas.h"
 #include "engine/engine_util_errmem.h"
 #include "engine/engine_util_misc.h"
@@ -526,9 +524,12 @@ static int isknown(const mjtNum* points, const mjtNum x[3], int cnt) {
 
 
 // adds candidate point to result
+// flipNormal: 0 = normal points INTO SDF (for mesh-SDF where SDF is g2)
+//             1 = normal points OUT of SDF (for flex-SDF where SDF is g1)
 static int addContact(mjtNum* points, mjContact* con, const mjtNum x[3],
                       const mjtNum pos2[3], const mjtNum quat2[4], mjtNum dist,
-                      int cnt, const mjModel* m, const mjSDF* s, mjData* d) {
+                      int cnt, const mjModel* m, const mjSDF* s, mjData* d,
+                      int flipNormal) {
   // check if there is a collision
   if (dist > 0 || isknown(points, x, cnt)) {
     return cnt;
@@ -539,7 +540,17 @@ static int addContact(mjtNum* points, mjContact* con, const mjtNum x[3],
   // compute normal in local coordinates
   mjtNum norm[3], vec[3];
   mjc_gradient(m, d, s, norm, x);
-  mju_scl3(norm, norm, -1);
+
+  // validate normal - skip if gradient is degenerate (zero or near-zero)
+  mjtNum norm_len = mju_normalize3(norm);
+  if (norm_len < mjMINVAL) {
+    return cnt;  // degenerate gradient, skip this contact
+  }
+
+  // normal direction: flipNormal=0 -> INTO SDF, flipNormal=1 -> OUT of SDF
+  if (!flipNormal) {
+    mju_scl3(norm, norm, -1);
+  }
 
   // construct contact
   con[cnt].dist = dist;
@@ -720,6 +731,106 @@ static int boxIntersect(const mjtNum bvh[6], const mjtNum offset[3],
   return mjc_distance(m, d, s, candidate) < r;
 }
 
+// Farthest point sampling (FPS) to select spatially diverse contacts
+// Returns indices of selected candidates in nselected, fills selected_indices array
+static int selectFPS(const mjtNum* candidate, const mjtNum* dist, int ncandidate,
+                     int* selected_indices, int max_select) {
+  if (ncandidate <= 0) return 0;
+
+  mjtByte selected[mjMAXCONPAIR] = {0};
+  mjtNum min_dist2[mjMAXCONPAIR];
+  for (int i = 0; i < ncandidate; i++) {
+    min_dist2[i] = mjMAXVAL;
+  }
+
+  // start with deepest penetrating contact
+  int best = 0;
+  mjtNum bestval = -dist[0];
+  for (int i = 1; i < ncandidate; i++) {
+    if (-dist[i] > bestval) {
+      bestval = -dist[i];
+      best = i;
+    }
+  }
+
+  // iteratively select contacts using FPS
+  int nselected = 0;
+  while (nselected < max_select && nselected < mjMAXCONPAIR && best >= 0) {
+    selected[best] = 1;
+    selected_indices[nselected] = best;
+    nselected++;
+
+    const mjtNum* bestpos = candidate + 3*best;
+
+    // find next farthest point
+    int nextbest = -1;
+    mjtNum nextbestdist = -1;
+    for (int i = 0; i < ncandidate; i++) {
+      if (selected[i]) continue;
+
+      mjtNum dx = candidate[3*i+0] - bestpos[0];
+      mjtNum dy = candidate[3*i+1] - bestpos[1];
+      mjtNum dz = candidate[3*i+2] - bestpos[2];
+      mjtNum d2 = dx*dx + dy*dy + dz*dz;
+      if (d2 < min_dist2[i]) {
+        min_dist2[i] = d2;
+      }
+      if (min_dist2[i] > nextbestdist) {
+        nextbestdist = min_dist2[i];
+        nextbest = i;
+      }
+    }
+    best = nextbest;
+  }
+
+  return nselected;
+}
+
+// Process triangle corners against SDF using Halton sampling + Frank-Wolfe.
+// Corners are assumed to already be in SDF local coordinates.
+// Adds penetrating candidates to candidate/dist arrays.
+static void processSdfCorners(const mjtNum corners[9], const mjModel* m, mjData* d,
+                              const mjSDF* sdf, int nstartpts,
+                              mjtNum* candidate, mjtNum* dist, int* ncandidate) {
+  mjtNum x[3], depth;
+
+  // stricter culling using triangle circumsphere
+  if (!triangleIntersect(corners, m, sdf, d)) {
+    return;
+  }
+
+  // sample multiple starting points using Halton sequence
+  for (int sp = 0; sp < nstartpts; sp++) {
+    if (*ncandidate >= mjMAXCONPAIR) break;
+
+    // barycentric coordinates from Halton sequence
+    mjtNum u = mju_Halton(sp + 1, 2);
+    mjtNum v = mju_Halton(sp + 1, 3);
+    if (u + v > 1) {
+      u = 1 - u;
+      v = 1 - v;
+    }
+    mjtNum b0 = 1 - u - v;
+    mjtNum b1 = u;
+    mjtNum b2 = v;
+
+    // starting point
+    x[0] = b0*corners[0] + b1*corners[3] + b2*corners[6];
+    x[1] = b0*corners[1] + b1*corners[4] + b2*corners[7];
+    x[2] = b0*corners[2] + b1*corners[5] + b2*corners[8];
+
+    depth = stepFrankWolfe(x, corners, 3, m, sdf, d);
+
+    // store candidate if penetration
+    if (depth < 0) {
+      int nc = *ncandidate;
+      mju_copy3(candidate + 3*nc, x);
+      dist[nc] = depth;
+      (*ncandidate)++;
+    }
+  }
+}
+
 // Context for inline face processing during BVH traversal
 typedef struct {
   const mjModel* m;
@@ -738,9 +849,8 @@ typedef struct {
 // process a single mesh face inline during BVH traversal
 static void processOneFace(int faceid, mjtByte* bvh_active, int node,
                            MeshSDFContext* ctx) {
-  mjtNum corners[9], x[3], depth;
+  mjtNum corners[9];
   const mjModel* m = ctx->m;
-  mjData* d = ctx->d;
   int face = ctx->faceadr + faceid;
 
   // transform triangle vertices to SDF local coordinates
@@ -754,85 +864,48 @@ static void processOneFace(int faceid, mjtByte* bvh_active, int node,
     mju_addTo3(corners + 3*v, ctx->offset);
   }
 
-  // stricter culling using triangle circumsphere
+  // use shared helper for Halton sampling + Frank-Wolfe
-  if (!triangleIntersect(corners, m, ctx->sdf, d)) {
+  processSdfCorners(corners, m, ctx->d, ctx->sdf, ctx->nstartpts,
-    return;
+                    ctx->candidate, ctx->dist, ctx->ncandidate);
-  }
 
-  if (bvh_active) bvh_active[node] = 1;
+  if (bvh_active && *(ctx->ncandidate) > 0) bvh_active[node] = 1;
-
-  // sample multiple starting points using Halton sequence
-  for (int sp = 0; sp < ctx->nstartpts; sp++) {
-    if (*(ctx->ncandidate) >= mjMAXCONPAIR) break;
-
-    // barycentric coordinates from Halton sequence
-    mjtNum u = mju_Halton(sp + 1, 2);
-    mjtNum v = mju_Halton(sp + 1, 3);
-    if (u + v > 1) {
-      u = 1 - u;
-      v = 1 - v;
-    }
-    mjtNum b0 = 1 - u - v;
-    mjtNum b1 = u;
-    mjtNum b2 = v;
-
-    // starting point
-    x[0] = b0*corners[0] + b1*corners[3] + b2*corners[6];
-    x[1] = b0*corners[1] + b1*corners[4] + b2*corners[7];
-    x[2] = b0*corners[2] + b1*corners[5] + b2*corners[8];
-
-    depth = stepFrankWolfe(x, corners, 3, m, ctx->sdf, d);
-
-    // store candidate if penetration
-    if (depth < 0) {
-      int nc = *(ctx->ncandidate);
-      mju_copy3(ctx->candidate + 3*nc, x);
-      ctx->dist[nc] = depth;
-      (*(ctx->ncandidate))++;
-    }
-  }
 }
 
-// tree vs sdf binary search with inline face processing
+// Callback type for leaf node processing during BVH traversal
-static void collideBVHInline(const mjModel* m, mjData* d, int g,
+// Returns 1 if node should be marked active (for visualization), 0 otherwise
-                             MeshSDFContext* ctx) {
+typedef int (*BVHLeafCallback)(int leaf_id, int node, void* ctx);
-  const int bvhadr = m->mesh_bvhadr[m->geom_dataid[g]];
 
-  // mesh has no BVH tree
+// Generic BVH traversal with callback for leaf processing
-  if (bvhadr < 0) {
+// bvh: pointer to BVH AABBs (6 floats per node: center xyz, half-size xyz)
-    return;
+// nodeid: leaf ID per node (-1 for intermediate nodes)
-  }
+// child: child indices (2 per node)
-
+// bvh_active: visualization array or NULL
-  const int* faceid = m->bvh_nodeid + bvhadr;
+// offset/rotation: transform for boxIntersect
-  const mjtNum* bvh = m->bvh_aabb + 6*bvhadr;
+// m, d, sdf: for boxIntersect
-  const int* child = m->bvh_child + 2*bvhadr;
+// callback: function called for each leaf node
-  mjtByte* bvh_active = m->vis.global.bvactive ? d->bvh_active + bvhadr : NULL;
+// ctx: user context passed to callback
-
+static void traverseBVH(const mjtNum* bvh, const int* nodeid, const int* child,
-  mj_markStack(d);
+                        mjtByte* bvh_active, const mjtNum* offset, const mjtNum* rotation,
-  int max_stack = m->mesh_bvhnum[m->geom_dataid[g]];
+                        const mjModel* m, mjData* d, const mjSDF* sdf,
-  struct CollideTreeArgs_ {
+                        BVHLeafCallback callback, void* ctx) {
-    int node;
+  int stack[64];
-  };
-  typedef struct CollideTreeArgs_ CollideTreeArgs;
-  CollideTreeArgs* stack = mjSTACKALLOC(d, max_stack, CollideTreeArgs);
   int nstack = 0;
-  stack[nstack].node = 0;
+  stack[nstack++] = 0;
-  nstack++;
 
   while (nstack) {
-    nstack--;
+    int node = stack[--nstack];
-    int node = stack[nstack].node;
 
-    // leaf node: process face inline
+    // leaf node: call callback
-    if (faceid[node] != -1) {
+    if (nodeid[node] != -1) {
-      if (boxIntersect(bvh + 6*node, ctx->offset, ctx->rotation, m, ctx->sdf, d)) {
+      if (boxIntersect(bvh + 6*node, offset, rotation, m, sdf, d)) {
-        processOneFace(faceid[node], bvh_active, node, ctx);
+        int active = callback(nodeid[node], node, ctx);
+        if (bvh_active && active) bvh_active[node] = 1;
       }
       continue;
     }
 
     // intermediate node: check bounding box
-    if (!boxIntersect(bvh + 6*node, ctx->offset, ctx->rotation, m, ctx->sdf, d)) {
+    if (!boxIntersect(bvh + 6*node, offset, rotation, m, sdf, d)) {
       continue;
     }
 
@@ -841,16 +914,24 @@ static void collideBVHInline(const mjModel* m, mjData* d, int g,
     // push children
     for (int i = 0; i < 2; i++) {
       if (child[2*node+i] != -1) {
-        if (nstack >= max_stack) mjERROR("BVH stack depth exceeded.");
+        if (nstack >= 64) {
-        stack[nstack].node = child[2*node+i];
+          mjERROR("BVH stack depth exceeded.");
-        nstack++;
+        }
+        stack[nstack++] = child[2*node+i];
       }
     }
   }
-
-  mj_freeStack(d);
 }
 
+// mesh face callback for traverseBVH
+static int meshFaceCallback(int face_id, int node, void* ctx) {
+  MeshSDFContext* mctx = (MeshSDFContext*)ctx;
+  int ncandidate_before = *(mctx->ncandidate);
+  processOneFace(face_id, NULL, node, mctx);
+  return *(mctx->ncandidate) > ncandidate_before;  // mark active if candidates added
+}
+
+
 //------------------------------ collision functions -----------------------------------------------
 
 // collision between a height field and a signed distance field
@@ -901,65 +982,36 @@ int mjc_MeshSDF(const mjModel* m, const mjData* d, mjContact* con, int g1, int g
   ctx.dist = dist;
   ctx.ncandidate = &ncandidate;
 
-  // binary tree search with inline face processing
+  // BVH traversal for mesh faces
-  collideBVHInline(m, (mjData*)d, g1, &ctx);
+  const int bvhadr = m->mesh_bvhadr[m->geom_dataid[g1]];
+  if (bvhadr >= 0) {
+    const int* nodeid = m->bvh_nodeid + bvhadr;
+    const mjtNum* bvh = m->bvh_aabb + 6*bvhadr;
+    const int* child = m->bvh_child + 2*bvhadr;
+    mjtByte* bvh_active = m->vis.global.bvactive ? d->bvh_active + bvhadr : NULL;
+
+    traverseBVH(bvh, nodeid, child, bvh_active, ctx.offset, ctx.rotation,
+                m, (mjData*)d, ctx.sdf, meshFaceCallback, &ctx);
+  }
 
   // if few candidates, add them all directly
   if (ncandidate <= mjMAXCONPAIR) {
     for (int i = 0; i < ncandidate; i++) {
       cnt = addContact(points, con, candidate + 3*i, pos2, sdf_quat,
-                       dist[i], cnt, m, &sdf, (mjData*)d);
+                       dist[i], cnt, m, &sdf, (mjData*)d, 0);
     }
     return cnt;
   }
 
   // farthest point sampling (FPS) for spatial coverage
-  mjtByte selected[mjMAXCONPAIR] = {0};
+  int selected_indices[mjMAXCONPAIR];
-  mjtNum min_dist2[mjMAXCONPAIR];
+  int nselected = selectFPS(candidate, dist, ncandidate, selected_indices, mjMAXCONPAIR);
-  for (int i = 0; i < ncandidate; i++) {
-    min_dist2[i] = mjMAXVAL;
-  }
 
-  // start with deepest penetrating contact
+  // add selected contacts
-  int best = 0;
+  for (int i = 0; i < nselected; i++) {
-  mjtNum bestval = -dist[0];
+    int idx = selected_indices[i];
-  for (int i = 1; i < ncandidate; i++) {
+    cnt = addContact(points, con, candidate + 3*idx, pos2, sdf_quat,
-    if (-dist[i] > bestval) {
+                     dist[idx], cnt, m, &sdf, (mjData*)d, 0);
-      bestval = -dist[i];
-      best = i;
-    }
-  }
-
-  // iteratively select contacts using FPS
-  int nselected = 0;
-  while (nselected < mjMAXCONPAIR && best >= 0) {
-    selected[best] = 1;
-    mjtNum* bestpos = candidate + 3*best;
-
-    // add the selected contact
-    cnt = addContact(points, con, bestpos, pos2, sdf_quat,
-                     dist[best], cnt, m, &sdf, (mjData*)d);
-    nselected++;
-
-    // find next farthest point
-    int nextbest = -1;
-    mjtNum nextbestdist = -1;
-    for (int i = 0; i < ncandidate; i++) {
-      if (selected[i]) continue;
-
-      mjtNum dx = candidate[3*i+0] - bestpos[0];
-      mjtNum dy = candidate[3*i+1] - bestpos[1];
-      mjtNum dz = candidate[3*i+2] - bestpos[2];
-      mjtNum d2 = dx*dx + dy*dy + dz*dz;
-      if (d2 < min_dist2[i]) {
-        min_dist2[i] = d2;
-      }
-      if (min_dist2[i] > nextbestdist) {
-        nextbestdist = min_dist2[i];
-        nextbest = i;
-      }
-    }
-    best = nextbest;
   }
 
   return cnt;
@@ -1092,7 +1144,7 @@ int mjc_SDF(const mjModel* m, const mjData* d, mjContact* con, int g1, int g2, m
 
     // contact point and normal - we use the midsurface where SDF1=SDF2 as zero level set
     sdf.type = mjSDFTYPE_MIDSURFACE;
-    cnt = addContact(contacts, con, x, pos2, quat2, dist, cnt, m, &sdf, (mjData*)d);
+    cnt = addContact(contacts, con, x, pos2, quat2, dist, cnt, m, &sdf, (mjData*)d, 0);
 
     // SHOULD NOT OCCUR
     if (cnt > mjMAXCONPAIR) {
@@ -1102,3 +1154,216 @@ int mjc_SDF(const mjModel* m, const mjData* d, mjContact* con, int g1, int g2, m
 
   return cnt;
 }
+
+// Context for flex element processing during BVH traversal
+typedef struct {
+  const mjModel* m;
+  mjData* d;
+  const mjSDF* sdf;
+  const mjtNum* offset;
+  const mjtNum* rotation;
+  const mjtNum* vertxpos;
+  int f;
+  int dim;
+  int nstartpts;
+  mjtNum* candidate;
+  mjtNum* dist;
+  int* elem_id;
+  int* ncandidate;
+} FlexSDFContext;
+
+// flex element callback for traverseBVH
+static int flexElemCallback(int elem_idx, int node, void* ctx) {
+  FlexSDFContext* fctx = (FlexSDFContext*)ctx;
+  const mjModel* m = fctx->m;
+  int dim = fctx->dim;
+
+  // get element vertex indices
+  const int* edata = m->flex_elem + m->flex_elemdataadr[fctx->f] + elem_idx*(dim+1);
+
+  // get triangle corners in world coordinates
+  mjtNum world_corners[9];
+  for (int v = 0; v < 3; v++) {
+    mju_copy3(world_corners + 3*v, fctx->vertxpos + 3*edata[v]);
+  }
+
+  // transform corners to SDF local coordinates
+  mjtNum corners[9];
+  for (int v = 0; v < 3; v++) {
+    mju_mulMatVec3(corners + 3*v, fctx->rotation, world_corners + 3*v);
+    mju_addTo3(corners + 3*v, fctx->offset);
+  }
+
+  // process triangle (per-element candidates)
+  mjtNum elem_candidate[3*mjMAXCONPAIR];
+  mjtNum elem_dist[mjMAXCONPAIR];
+  int elem_ncandidate = 0;
+  processSdfCorners(corners, m, fctx->d, fctx->sdf, fctx->nstartpts,
+                    elem_candidate, elem_dist, &elem_ncandidate);
+
+  // keep only deepest penetrating candidate from this element
+  if (elem_ncandidate > 0) {
+    int best_i = 0;
+    for (int i = 1; i < elem_ncandidate; i++) {
+      if (elem_dist[i] < elem_dist[best_i]) {
+        best_i = i;
+      }
+    }
+    // add to global candidate list
+    int nc = *(fctx->ncandidate);
+    if (nc < mjMAXCONPAIR) {
+      mju_copy3(fctx->candidate + 3*nc, elem_candidate + 3*best_i);
+      fctx->dist[nc] = elem_dist[best_i];
+      fctx->elem_id[nc] = elem_idx;
+      (*(fctx->ncandidate))++;
+      return 1;  // mark active
+    }
+  }
+  return 0;
+}
+
+
+int mjc_FlexSDF(const mjModel* m, mjData* d, mjContact* con,
+                int g, int f, mjtNum margin) {
+  // g = SDF geom, f = flex
+  int dim = m->flex_dim[f];
+
+  // only support dim==2 (triangular elements)
+  if (dim != 2) {
+    return 0;
+  }
+
+  mjtNum points[3*mjMAXCONPAIR], dist[mjMAXCONPAIR], candidate[3*mjMAXCONPAIR];
+  int elem_id[mjMAXCONPAIR];  // track which element generated each candidate
+  int cnt = 0, ncandidate = 0;
+
+  // get SDF info (once for entire flex)
+  int instance = m->geom_plugin[g];
+  const mjpPlugin* sdf_ptr = instance == -1 ? NULL : mjc_getSDF(m, g);
+  instance = instance == -1 ? m->geom_dataid[g] : instance;
+  mjtGeom geomtype = mjGEOM_SDF;
+
+  mjSDF sdf;
+  sdf.id = &instance;
+  sdf.type = mjSDFTYPE_SINGLE;
+  sdf.plugin = &sdf_ptr;
+  sdf.geomtype = &geomtype;
+
+  // get SDF pose (once for entire flex)
+  mjtNum sdf_quat[4];
+  const mjtNum* sdf_pos = d->geom_xpos + 3*g;
+  const mjtNum* sdf_mat = d->geom_xmat + 9*g;
+  mju_mat2Quat(sdf_quat, sdf_mat);
+
+  // compute world-to-SDF transform (once for entire flex)
+  mjtNum rotation[9], offset[3];
+  mjtNum world_origin[3] = {0, 0, 0}, world_quat[4] = {1, 0, 0, 0};
+  mapPose(world_origin, world_quat, sdf_pos, sdf_quat, offset, rotation);
+
+  // get flex element and vertex data
+  const mjtNum* vertxpos = d->flexvert_xpos + 3*m->flex_vertadr[f];
+  int nstartpts = mju_max(1, m->opt.sdf_initpoints);
+  int elemnum = m->flex_elemnum[f];
+
+  // flex has no BVH tree: iterate all elements directly
+  const int bvhadr = m->flex_bvhadr[f];
+  if (bvhadr < 0) {
+    mjtNum elem_candidate[3*mjMAXCONPAIR];
+    mjtNum elem_dist[mjMAXCONPAIR];
+    for (int e = 0; e < elemnum; e++) {
+      const int* edata = m->flex_elem + m->flex_elemdataadr[f] + e*(dim+1);
+
+      // get triangle corners in world coordinates
+      mjtNum world_corners[9];
+      for (int v = 0; v < 3; v++) {
+        mju_copy3(world_corners + 3*v, vertxpos + 3*edata[v]);
+      }
+
+      // transform corners to SDF local coordinates
+      mjtNum corners[9];
+      for (int v = 0; v < 3; v++) {
+        mju_mulMatVec3(corners + 3*v, rotation, world_corners + 3*v);
+        mju_addTo3(corners + 3*v, offset);
+      }
+
+      // process triangle using shared helper (per-element candidates)
+      int elem_ncandidate = 0;
+      processSdfCorners(corners, m, d, &sdf, nstartpts, elem_candidate, elem_dist, &elem_ncandidate);
+
+      // add contacts for this element immediately (per-element points tracking)
+      mjtNum elem_points[3*mjMAXCONPAIR];
+      int elem_cnt = 0;
+      for (int i = 0; i < elem_ncandidate && cnt < mjMAXCONPAIR; i++) {
+        int old_elem_cnt = elem_cnt;
+        elem_cnt = addContact(elem_points, con + cnt, elem_candidate + 3*i, sdf_pos, sdf_quat,
+                              elem_dist[i], elem_cnt, m, &sdf, d, 1);
+        // set element ID for successfully added contact
+        if (elem_cnt > old_elem_cnt) {
+          con[cnt].elem[1] = e;
+          cnt++;
+        }
+      }
+    }
+    return cnt;
+  }
+
+  // have BVH: use tree traversal for efficiency
+  const int* elemid = m->bvh_nodeid + bvhadr;
+  // flexes use dynamic AABBs (vertices move during simulation)
+  const int nbvhstatic = m->nbvhstatic;
+  const mjtNum* bvh = d->bvh_aabb_dyn + 6*(bvhadr - nbvhstatic);
+  const int* child = m->bvh_child + 2*bvhadr;
+  mjtByte* bvh_active = m->vis.global.bvactive ? d->bvh_active + bvhadr : NULL;
+
+  // set up context for flex element processing
+  FlexSDFContext fctx;
+  fctx.m = m;
+  fctx.d = d;
+  fctx.sdf = &sdf;
+  fctx.offset = offset;
+  fctx.rotation = rotation;
+  fctx.vertxpos = vertxpos;
+  fctx.f = f;
+  fctx.dim = dim;
+  fctx.nstartpts = nstartpts;
+  fctx.candidate = candidate;
+  fctx.dist = dist;
+  fctx.elem_id = elem_id;
+  fctx.ncandidate = &ncandidate;
+
+  traverseBVH(bvh, elemid, child, bvh_active, offset, rotation,
+              m, d, &sdf, flexElemCallback, &fctx);
+
+
+  // if few candidates, add them all directly
+  if (ncandidate <= mjMAXCONPAIR) {
+    for (int i = 0; i < ncandidate; i++) {
+      int old_cnt = cnt;
+      cnt = addContact(points, con, candidate + 3*i, sdf_pos, sdf_quat,
+                       dist[i], cnt, m, &sdf, d, 1);
+      // set element ID for successfully added contact
+      if (cnt > old_cnt) {
+        con[old_cnt].elem[1] = elem_id[i];
+      }
+    }
+    return cnt;
+  }
+
+  // farthest point sampling (FPS) for spatial coverage
+  int selected_indices[mjMAXCONPAIR];
+  int nselected = selectFPS(candidate, dist, ncandidate, selected_indices, mjMAXCONPAIR);
+
+  // add selected contacts
+  for (int i = 0; i < nselected; i++) {
+    int idx = selected_indices[i];
+    int old_cnt = cnt;
+    cnt = addContact(points, con, candidate + 3*idx, sdf_pos, sdf_quat,
+                     dist[idx], cnt, m, &sdf, d, 1);
+    // set element ID for successfully added contact
+    if (cnt > old_cnt) {
+      con[old_cnt].elem[1] = elem_id[idx];
+    }
+  }
+
+  return cnt;
+}

src/engine/engine_collision_sdf.h

@@ -44,6 +44,10 @@ int mjc_MeshSDF(const mjModel* m, const mjData* d, mjContact* con, int g1, int g
 // collision between two signed distance fields
 int mjc_SDF(const mjModel* m, const mjData* d, mjContact* con, int g1, int g2, mjtNum margin);
 
+// collision between entire flex and SDF geom (batched processing)
+int mjc_FlexSDF(const mjModel* m, mjData* d, mjContact* con,
+                int g, int f, mjtNum margin);
+
 #ifdef __cplusplus
 }
 #endif

test/engine/engine_collision_sdf_test.cc

@@ -81,5 +81,71 @@ TEST_F(SdfTest, SdfPrimitive) {
   mj_deleteModel(model);
 }
 
+static constexpr char kFlexSdfModel[] = R"(
+<mujoco>
+  <extension>
+    <plugin plugin="mujoco.sdf.torus">
+      <instance name="torus">
+        <config key="radius1" value="0.35"/>
+        <config key="radius2" value="0.15"/>
+      </instance>
+    </plugin>
+  </extension>
+
+  <asset>
+    <mesh name="torus">
+      <plugin instance="torus"/>
+    </mesh>
+  </asset>
+
+  <worldbody>
+    <body pos="0 0 -0.3">
+      <geom type="sdf" mesh="torus">
+        <plugin instance="torus"/>
+      </geom>
+    </body>
+    <body name="flex">
+      <flexcomp name="test" type="grid" count="3 3 1" spacing=".2 .2 .2" dim="2">
+        <elasticity young="1e4"/>
+        <contact selfcollide="none" internal="false"/>
+      </flexcomp>
+    </body>
+  </worldbody>
+</mujoco>
+)";
+
+TEST_F(SdfTest, FlexSdfCollision) {
+  char error[1024];
+  mjModel* model = LoadModelFromString(kFlexSdfModel, error, sizeof(error));
+  ASSERT_THAT(model, NotNull()) << error;
+  mjData* data = mj_makeData(model);
+  ASSERT_THAT(data, NotNull());
+
+  // check there is at least one flex
+  ASSERT_GT(model->nflex, 0);
+
+  // simulate for a few steps to let flex fall onto torus
+  for (int i = 0; i < 100; i++) {
+    mj_step(model, data);
+  }
+
+  // should have contacts between flex and SDF
+  EXPECT_GT(data->ncon, 0) << "Expected flex-SDF contacts";
+
+  // check that contacts involve the flex
+  bool has_flex_contact = false;
+  for (int i = 0; i < data->ncon; i++) {
+    if (data->contact[i].flex[0] >= 0 || data->contact[i].flex[1] >= 0) {
+      has_flex_contact = true;
+      break;
+    }
+  }
+  EXPECT_TRUE(has_flex_contact) << "Expected at least one flex contact";
+
+  mj_deleteData(data);
+  mj_deleteModel(model);
+}
+
 }  // namespace
 }  // namespace mujoco
+