[Newton] Add friction randomization event #3499

Milad-Rakhsha-NV · 2025-09-18T19:22:35Z

Description

Please include a summary of the change and which issue is fixed. Please also include relevant motivation and context.
List any dependencies that are required for this change.

Fixes # (issue)

Type of change

Bug fix (non-breaking change which fixes an issue)
New feature (non-breaking change which adds functionality)
Breaking change (existing functionality will not work without user modification)
Documentation update

Screenshots

Please attach before and after screenshots of the change if applicable.

Checklist

I have read and understood the contribution guidelines
I have run the pre-commit checks with ./isaaclab.sh --format
I have made corresponding changes to the documentation
My changes generate no new warnings
I have added tests that prove my fix is effective or that my feature works
I have updated the changelog and the corresponding version in the extension's config/extension.toml file
I have added my name to the CONTRIBUTORS.md or my name already exists there

Mayankm96 · 2025-09-22T07:13:47Z

source/isaaclab/isaaclab/envs/mdp/events.py

 class randomize_rigid_body_material(ManagerTermBase):
    def __init__(self, cfg: EventTermCfg, env: ManagerBasedEnv):
-        raise NotImplementedError("Not implemented")
+        """Initialize the term.


do we need this concept of bucketing for Newton?

I don't think so, @Milad-Rakhsha-NV there is no limit of materials that can be applied in Newton right? They should just be parameters for MJWarp to process.

yeah there is no limit but since I wanted to keep API consistency with the physX-based lab I just did the same thing here;
we can keep the same API and just not use some parameters internally (e.g. num_buckets/restitution which aren't used)
or we can just have a nen API for Newton (which I wouldn't prefer for obvious reasons)

AntoineRichard

Thanks Milad for the PR! It looks great though there are some concerns around the sampling scheme. Could you take a look? Thanks!

AntoineRichard · 2025-09-22T07:36:25Z

source/isaaclab/isaaclab/envs/mdp/events.py

 class randomize_rigid_body_material(ManagerTermBase):
    def __init__(self, cfg: EventTermCfg, env: ManagerBasedEnv):
-        raise NotImplementedError("Not implemented")
+        """Initialize the term.


I don't think so, @Milad-Rakhsha-NV there is no limit of materials that can be applied in Newton right? They should just be parameters for MJWarp to process.

AntoineRichard · 2025-09-22T07:37:41Z

source/isaaclab/isaaclab/envs/mdp/events.py

+        # obtain number of shapes per body (needed for indexing the material properties correctly)
+        # note: this is a workaround since the Articulation does not provide a direct way to obtain the number of shapes
+        #  per body. We use the physics simulation view to obtain the number of shapes per body.
+        self.num_shapes_per_body = None
+        if isinstance(self.asset, Articulation) and self.asset_cfg.body_ids != slice(None):
+            self.num_shapes_per_body = []
+            for shapes in self.asset.root_newton_view.body_shapes:
+                self.num_shapes_per_body.append(len(shapes))


Should we move this to the articulation? Newton does provide a body/shape mapping.

AntoineRichard · 2025-09-22T07:38:36Z

source/isaaclab/isaaclab/envs/mdp/events.py

+        self.material_buckets = math_utils.sample_uniform(
+            ranges[:, 0], ranges[:, 1], (num_buckets, 3), device=self.asset.device
+        )
+
+        # ensure dynamic friction is always less than static friction
+        make_consistent = cfg.params.get("make_consistent", False)
+        if make_consistent:
+            self.material_buckets[:, 1] = torch.min(self.material_buckets[:, 0], self.material_buckets[:, 1])


if there is no limit, we should sample these values dynamically on every call. Could you change the behavior to that?

AntoineRichard · 2025-09-22T07:38:55Z

source/isaaclab/isaaclab/envs/mdp/events.py

+        # obtain number of shapes per body (needed for indexing the material properties correctly)
+        # note: this is a workaround since the Articulation does not provide a direct way to obtain the number of shapes
+        #  per body. We use the physics simulation view to obtain the number of shapes per body.


Don't forget to update the comments! :)

AntoineRichard · 2025-09-22T07:39:17Z

source/isaaclab/isaaclab/envs/mdp/events.py

+            material_mu[env_ids, :] = material_samples[:, :, 0]
+
+        # apply to simulation
+        mask = torch.zeros((env.scene.num_envs,), dtype=torch.bool, device=env.device)


This should be cached.

AntoineRichard · 2025-09-22T07:41:52Z

source/isaaclab/isaaclab/envs/mdp/events.py

+        # retrieve material buffer from the physics simulation
+        # materials = self.asset.root_physx_view.get_material_properties()
+        material_mu = wp.to_torch(
+            self.asset.root_newton_view.get_attribute("shape_material_mu", self.asset.root_newton_model)


These should be cached. You could save the "default" friction on init or on the first call. Also, if the tensor is contiguous then there is no need to clone as the values could be directly changed in place without the need to call the set function.

AntoineRichard · 2025-09-22T07:43:40Z

source/isaaclab/isaaclab/envs/mdp/events.py

+        if make_consistent:
+            self.material_buckets[:, 1] = torch.min(self.material_buckets[:, 0], self.material_buckets[:, 1])

    def __call__(


Should we consider using warp for this? @Mayankm96 do you think it would make sense. Using warp would avoid copies and in general make the update much faster as we could do this whole function in a couple of lines. It's not really visible from the user POV so I'd say it's fine.

for mjwarp actually this is not used since there is a single friction value (not including the spinning/rolling frictions) that can be set to the solver.

AntoineRichard · 2025-09-22T07:43:53Z

source/isaaclab/isaaclab/envs/mdp/events.py

+        # retrieve material buffer from the physics simulation
+        # materials = self.asset.root_physx_view.get_material_properties()


These would need to removed?

AntoineRichard · 2025-09-22T07:44:20Z

source/isaaclab/isaaclab/envs/mdp/events.py

+        bucket_ids = torch.randint(0, num_buckets, (len(env_ids), total_num_shapes), device=self.asset.device)
+        material_samples = self.material_buckets[bucket_ids]
+        # print("material_mu before update:\n", material_mu.shape, material_mu)
+        # update material buffer with new samples
+        if self.num_shapes_per_body is not None:
+            # sample material properties from the given ranges
+            for body_id in self.asset_cfg.body_ids:
+                # obtain indices of shapes for the body
+                start_idx = sum(self.num_shapes_per_body[:body_id])
+                end_idx = start_idx + self.num_shapes_per_body[body_id]
+                # assign the new materials
+                # material samples are of shape: num_env_ids x total_num_shapes x 3
+                material_mu[env_ids, start_idx:end_idx] = material_samples[:, start_idx:end_idx, 0]
+        else:
+            # assign all the materials
+            material_mu[env_ids, :] = material_samples[:, :, 0]


Update to use direct sampling if possible!

AntoineRichard · 2025-09-22T07:44:34Z

source/isaaclab/isaaclab/envs/mdp/events.py

+        # material_mu = wp.to_torch(self.asset.root_newton_view.get_attribute("shape_material_mu", self.asset.root_newton_model)).clone()
+        # print("material_mu after update:\n", material_mu.shape, material_mu)


This needs to be removed

…randomization

Signed-off-by: Milad-Rakhsha <[email protected]>

friction randomization

6d69d71

Milad-Rakhsha-NV requested review from kellyguo11 and Mayankm96 as code owners September 18, 2025 19:22

Milad-Rakhsha-NV requested a review from AntoineRichard September 18, 2025 19:23

Mayankm96 reviewed Sep 22, 2025

View reviewed changes

AntoineRichard requested changes Sep 22, 2025

View reviewed changes

AntoineRichard changed the title ~~Added friction randomization event~~ [Newton] Add friction randomization event Sep 22, 2025

Merge remote-tracking branch 'origin/dev/newton' into milad/friction-…

c3220cb

…randomization

Milad-Rakhsha-NV requested a review from jtigue-bdai as a code owner September 22, 2025 22:45

Milad-Rakhsha-NV force-pushed the milad/friction-randomization branch 4 times, most recently from 93ac402 to be1f27a Compare September 22, 2025 23:48

Addressed the code review comments

81490c2

Signed-off-by: Milad-Rakhsha <[email protected]>

Milad-Rakhsha-NV force-pushed the milad/friction-randomization branch from be1f27a to 81490c2 Compare September 23, 2025 01:40

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

[Newton] Add friction randomization event #3499

[Newton] Add friction randomization event #3499

Milad-Rakhsha-NV commented Sep 18, 2025

Uh oh!

Mayankm96 Sep 22, 2025

Uh oh!

AntoineRichard Sep 22, 2025

Uh oh!

Milad-Rakhsha-NV Sep 22, 2025

Uh oh!

AntoineRichard left a comment

Uh oh!

AntoineRichard Sep 22, 2025

Uh oh!

AntoineRichard Sep 22, 2025

Uh oh!

AntoineRichard Sep 22, 2025

Uh oh!

AntoineRichard Sep 22, 2025

Uh oh!

AntoineRichard Sep 22, 2025

Uh oh!

AntoineRichard Sep 22, 2025

Uh oh!

AntoineRichard Sep 22, 2025

Uh oh!

Milad-Rakhsha-NV Sep 22, 2025

Uh oh!

AntoineRichard Sep 22, 2025

Uh oh!

AntoineRichard Sep 22, 2025

Uh oh!

AntoineRichard Sep 22, 2025

Uh oh!

Uh oh!

		# retrieve material buffer from the physics simulation
		# materials = self.asset.root_physx_view.get_material_properties()

		# material_mu = wp.to_torch(self.asset.root_newton_view.get_attribute("shape_material_mu", self.asset.root_newton_model)).clone()
		# print("material_mu after update:\n", material_mu.shape, material_mu)

[Newton] Add friction randomization event #3499

Are you sure you want to change the base?

[Newton] Add friction randomization event #3499

Conversation

Milad-Rakhsha-NV commented Sep 18, 2025

Description

Type of change

Screenshots

Checklist

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

AntoineRichard left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!