newton/examples/assets/cartpole.usda (2)

92-113: Nested Cube under pole2 is misnamed "pole1" — rename for consistency

The nested Gprim under Xform "pole2" is currently named "pole1". While unique by full path, this is confusing and error-prone.

Apply this rename:

-        def Cube "pole1" (
+        def Cube "pole2" (
             prepend apiSchemas = ["PhysicsCollisionAPI"]
         )

---

129-138: Verify hinge anchor placement for cart-pole joint

localPos0 on the cart is set to (0.55, 0.0, 0), which places the hinge well outside the cart’s 0.2 length along X. Typically, the pole should hinge at the top center of the cart, i.e., around (0, 0, +half_cart_height). Consider moving the anchor to the cart’s top face.

Proposed adjustment:

-        point3f physics:localPos0 = (0.55, 0.0, 0)
+        # Hinge at the top center of the cart (cart scale along Z is 0.2 => half-height ~0.1)
+        point3f physics:localPos0 = (0.0, 0.0, 0.1)

Please verify visually/simulation-wise and tweak as needed to avoid initial interpenetration.

newton/examples/example_cartpole.py (3)

19-22: Update comment to reflect USD-based import (not URDF)

The example now loads from USD. Adjust the comment to prevent confusion.

-# from a URDF using newton.ModelBuilder().
+# from a USD stage using newton.ModelBuilder().

---

42-47: Switch to parse_usd is correct; consider applying stage up-axis automatically

The call looks good. Optionally pass apply_up_axis_from_stage=True to honor the stage’s upAxis (Z) in environments where the default differs.

-        newton.utils.parse_usd(
+        newton.utils.parse_usd(
             newton.examples.get_asset("cartpole.usda"),
             articulation_builder,
-            enable_self_collisions=False,
-            collapse_fixed_joints=True,
+            enable_self_collisions=False,
+            collapse_fixed_joints=True,
+            apply_up_axis_from_stage=True,
         )

---

63-65: Assumes exactly three joints per env; consider making this resilient to asset changes

Setting builder.joint_q[-3:] works with the current asset (prismatic + two revolutes). If the USD changes joint count/order, this will silently misinitialize. If available, prefer targeting joints by name/path via the importer’s mappings or by capturing joint indices at import time.

I can help wire this to importer-returned mappings or generate a small utility to resolve joint indices by path. Want me to draft that?

Configuration used: .coderabbit.yml
Review profile: CHILL
Plan: Pro

💡 Knowledge Base configuration:

• MCP integration is disabled by default for public repositories
• Jira integration is disabled by default for public repositories
• Linear integration is disabled by default for public repositories

You can enable these sources in your CodeRabbit configuration.

newton/examples/assets/cartpole.usda (1)

47-66: Mass/inertia moved to the body prim; color moved to the geom — exactly right

This addresses prior feedback about placing mass properties on the rigid body prim (Xform with PhysicsMassAPI) and visual color on the mesh prim. Thanks for fixing this.

newton/examples/example_mujoco.py (1)

148-153: Minor: consider clarifying the “root pose” comment to “initial joint state”

Given this is a fixed-base articulation, the values immediately after parsing represent the cart prismatic DOF and the two pole angles, not a free-base root pose. Renaming the comment will prevent future confusion.

newton/examples/assets/cartpole.usda (3)

26-45: Articulation root and rail structure are reasonable

Using PhysicsArticulationRootAPI with a fixed joint to world via the rail is valid. The rail being an Xform with a Collision-only child is consistent with a static base.

If you intend the rail to be fully static, you can drop PhysicsRigidBodyAPI from the rail Xform to make that intent explicit and rely on the FixedJoint-to-world. Not critical, but can reduce ambiguity for consumers that infer dynamics from APIs.

---

55-58: Nit: replace tabs with spaces for consistency

There are a few lines indented with tabs; the rest uses spaces. While USD is whitespace-agnostic, mixed indentation can trip formatters and reviews.

Apply:

-		float3 physics:diagonalInertia = (0.1, 0.1, 0.1)
-		float physics:mass = 1
-		quatf physics:principalAxes = (1, 0, 0, 0)
+        float3 physics:diagonalInertia = (0.1, 0.1, 0.1)
+        float physics:mass = 1
+        quatf physics:principalAxes = (1, 0, 0, 0)
@@
-			color3f[] primvars:displayColor = [(0.3, 0.5, 0.7)]
+            color3f[] primvars:displayColor = [(0.3, 0.5, 0.7)]

Also applies to: 64-64

---

128-137: Cart–pole joint frames: please double-check anchor placement

The pole rotates about X, which is fine, but the cart anchor at (0.55, 0, 0) seems offset along X far beyond the cart’s half-extent (scale X=0.2). If this isn’t by design, consider moving the anchor near the cart’s top center and adjusting localPos1 accordingly.

Configuration used: .coderabbit.yml
Review profile: CHILL
Plan: Pro

💡 Knowledge Base configuration:

• MCP integration is disabled by default for public repositories
• Jira integration is disabled by default for public repositories
• Linear integration is disabled by default for public repositories

You can enable these sources in your CodeRabbit configuration.

Learnt from: nvlukasz
PR: newton-physics/newton#519
File: newton/utils.py:30-33
Timestamp: 2025-08-12T18:07:17.267Z
Learning: In newton/_src/utils/import_usd.py, the PXR/USD modules are lazily loaded using try-except blocks within the functions rather than at module import time, making it safe to import parse_usd at the module level without causing import-time failures for users without USD dependencies.

Learnt from: nvlukasz
PR: newton-physics/newton#519
File: newton/utils.py:30-33
Timestamp: 2025-08-12T18:07:17.267Z
Learning: In newton/_src/utils/import_usd.py, the PXR/USD modules are lazily loaded using try-except blocks within the functions (lines 109-111 and 1289-1291) rather than at module import time, making it safe to import parse_usd at the module level without causing import-time failures for users without USD dependencies.

newton/examples/assets/cartpole.usda (1)

47-66: Mass properties correctly moved to the body Xform; color authored on the mesh

This addresses the earlier review about mass/inertia being on the shape. The body now carries PhysicsMassAPI attributes and the Cube carries displayColor — parsers will pick these up as intended.

newton/examples/assets/cartpole.usda (3)

55-58: Nit: mixed tabs/spaces — standardize indentation

The lines authoring physics:diagonalInertia/mass/principalAxes and primvars:displayColor use tabs, while the rest of the file uses spaces. Standardize to spaces to avoid diffs/noise across editors.

Apply this whitespace-only diff:

-		float3 physics:diagonalInertia = (0.1, 0.1, 0.1)
-		float physics:mass = 1
-		quatf physics:principalAxes = (1, 0, 0, 0)
+        float3 physics:diagonalInertia = (0.1, 0.1, 0.1)
+        float physics:mass = 1
+        quatf physics:principalAxes = (1, 0, 0, 0)
@@
-			color3f[] primvars:displayColor = [(0.3, 0.5, 0.7)]
+            color3f[] primvars:displayColor = [(0.3, 0.5, 0.7)]

Also applies to: 64-64

---

128-137: Hinge anchor likely misplaced (localPos0 is along X, not Z/top of cart)

For a standard cart-pole, the hinge sits on the cart’s top face (Z half-extent), not offset along X. With cart scale z = 0.2, half-height is 0.1. Recommend moving the hinge anchor on the cart to z = +0.1.

Proposed fix:

 def PhysicsRevoluteJoint "cartPoleJoint"
 {
     rel physics:body0 = </cartPole/cart>
     rel physics:body1 = </cartPole/pole1>
     uniform token physics:axis = "X"
-    point3f physics:localPos0 = (0.55, 0.0, 0)
+    point3f physics:localPos0 = (0, 0, 0.1)
     point3f physics:localPos1 = (0, 0, -0.5)
     quatf physics:localRot0 = (1, 0, 0, 0)
     quatf physics:localRot1 = (1, 0, 0, 0)
 }

Please verify in the viewer that the pole bottom (localPos1 = (0,0,-0.5)) now meets the cart top without gaps/overlap, and that the swing plane remains about the X axis.

---

55-58: Consider letting the importer compute inertia from mass and geometry

The diagonalInertia/principalAxes are currently hard-coded and may not match the actual geometry scales and masses (e.g., poles at 0.25 kg with 1 m length). To avoid incorrect dynamics, either compute physically correct box inertia or omit these attributes so the importer computes inertia from shapes.

Option A (simplest): remove explicit inertia/axes so they’re derived:

@@
-        float3 physics:diagonalInertia = (0.1, 0.1, 0.1)
-        float physics:mass = 1
-        quatf physics:principalAxes = (1, 0, 0, 0)
+        float physics:mass = 1
@@
-        float3 physics:diagonalInertia = (0.1, 0.1, 0.1)
-        float physics:mass = 0.25
-        quatf physics:principalAxes = (1, 0, 0, 0)
+        float physics:mass = 0.25
@@
-        float3 physics:diagonalInertia = (0.1, 0.1, 0.1)
-        float physics:mass = 0.25
-        quatf physics:principalAxes = (1, 0, 0, 0)
+        float physics:mass = 0.25

Option B: if you want explicit inertia, I can provide the closed-form box inertia values matching your scales and masses.

Also applies to: 76-80, 99-103

newton/examples/example_mujoco.py (1)

148-153: Recommend aligning builder’s up‐axis with the USD stage

The USD file declares upAxis="Z", whereas ModelBuilder’s default is Y-up. To avoid unexpected rotations when loading Z-up stages, enable apply_up_axis_from_stage in the parse_usd call:

 newton.utils.parse_usd(
     newton.examples.get_asset("cartpole.usda"),
     articulation_builder,
     enable_self_collisions=False,
     collapse_fixed_joints=True,
+    apply_up_axis_from_stage=True,
 )

(optional — improves robustness when importing Z-up USD assets)

Configuration used: .coderabbit.yml
Review profile: CHILL
Plan: Pro

💡 Knowledge Base configuration:

• MCP integration is disabled by default for public repositories
• Jira integration is disabled by default for public repositories
• Linear integration is disabled by default for public repositories

You can enable these sources in your CodeRabbit configuration.