fix: multi-world particle BVH indexing

[Idate96]

Problem

SensorTiledCamera builds the particle BVH over the global particle arrays (particles_position/particles_radius) and uses BVH groups to restrict queries to a given world.

In newton/_src/sensors/warp_raytrace/ray_cast.py, particle hits were indexed as if wp.bvh_query_next() returned a world-local index (subtracting world_index * bvh_particles_size).

For multi-world particle scenes this produces invalid indices (and can trigger CUDA illegal memory access) and/or incorrect depth images for world_index > 0.

Fix

Treat the BVH hit index returned by wp.bvh_query_next() as a global index into the particle arrays, and remove the per-world offset in:

• closest_hit_particles()
• first_hit_particles()

Test

Add a minimal regression test that renders a tiny particle grid in multiple worlds and asserts the per-world depth images match (within a small fp32 tolerance):

• newton/tests/test_sensor_tiled_camera_particles_multiworld.py

The test uses the existing device parametrization helpers, so it runs on CPU and CUDA.

Summary by CodeRabbit

• Bug Fixes

  • Fixed particle ray-casting index handling to correct depth results in sensor rendering across multiple worlds.

• Tests

  • Added a regression test validating multi-world particle sensor depth consistency across transforms and devices.

[linux-foundation-easycla]

The committers listed above are authorized under a signed CLA.

• ✅ login: Idate96 / name: Lorenzo Terenzi (102837c, 71f669f, 9d1059f)

[coderabbitai]

📝 Walkthrough

Walkthrough

Refactored particle index handling in ray casting to use BVH query indices directly as global indices into particle arrays, removing local remapping. Added a regression test that builds a 4-world particle scene and verifies SensorTiledCamera depth images are identical across translated worlds.

Changes

Cohort / File(s)	Summary
Ray casting index handling newton/_src/sensors/warp_raytrace/ray_cast.py	Simplified particle traversal in closestHit and firstHit to treat indices returned by wp.bvh_query_next() as global indices into particles_position/particles_radius. Removed prior global->local remapping and added comments documenting the global-index assumption and BVH construction over global particle arrays.
Multi-world camera consistency test newton/tests/test_sensor_tiled_camera_particles_multiworld.py	New regression test module that constructs a multi-world particle model with per-world translations and verifies SensorTiledCamera depth-image consistency across 4 worlds. Adds helper _build_multiworld_particle_model, device-parametrized test test_sensor_tiled_camera_multiworld_particles_consistent, and a unittest harness class.

Estimated code review effort

🎯 3 (Moderate) | ⏱️ ~20 minutes

Possibly related PRs

• Adding newton.sensors.TiledCameraSensor with warp-raytrace #1044: Directly modifies newton/_src/sensors/warp_raytrace/ray_cast.py with particle-index handling changes in the same ray casting implementation.
• Add particle raycasting support to RaycastSensor and related tests #1056: Modifies particle raycasting traversal logic, relating to the BVH query index interpretation changes in this PR.

Suggested reviewers

• eric-heiden

🚥 Pre-merge checks | ✅ 3 | ❌ 1

❌ Failed checks (1 warning)

Check name	Status	Explanation	Resolution
Docstring Coverage	⚠️ Warning	Docstring coverage is 50.00% which is insufficient. The required threshold is 80.00%.	Write docstrings for the functions missing them to satisfy the coverage threshold.

✅ Passed checks (3 passed)

Check name	Status	Explanation
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.
Title check	✅ Passed	The title 'fix: multi-world particle BVH indexing' directly and concisely describes the main change: correcting BVH index handling for multi-world particle scenes.
Merge Conflict Detection	✅ Passed	✅ No merge conflicts detected when merging into main

_{✏️ Tip: You can configure your own custom pre-merge checks in the settings.}

✨ Finishing touches

• 📝 Generate docstrings

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Post copyable unit tests in a comment

---

No actionable comments were generated in the recent review. 🎉

🧹 Recent nitpick comments

newton/tests/test_sensor_tiled_camera_particles_multiworld.py (1)

27-72: Well-structured helper — clean and deterministic.

The builder logic is clear and the docstring explains intent well. Two tiny nits, both entirely optional:

1. Line 66: int(worlds) is redundant since worlds is already typed as int.
2. Lines 38/40: Ruff TRY003 suggests moving long exception messages into custom exception classes, but for a test helper this is fine.

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands and usage tips.}

[coderabbitai]

Actionable comments posted: 1

🤖 Fix all issues with AI agents

In `@newton/tests/test_sensor_tiled_camera_particles_multiworld.py`:
- Around line 1-2: Update the SPDX header in the new file by changing the
copyright year from 2025 to 2026: modify the two header lines starting with
"SPDX-FileCopyrightText" and "SPDX-License-Identifier" so the copyright year
reads 2026 (i.e., replace "Copyright (c) 2025" with "Copyright (c) 2026") to
match the current-year requirement.

🧹 Nitpick comments (1)

newton/tests/test_sensor_tiled_camera_particles_multiworld.py (1)

98-111: Inconsistent quaternion construction — prefer wp.quat_identity() for clarity.

Line 69 uses wp.quat_identity() but line 98 constructs the identity quaternion manually as wp.quatf(0.0, 0.0, 0.0, 1.0). Using the same helper would be clearer and less error-prone.

Proposed fix

-    cam_quat = wp.quatf(0.0, 0.0, 0.0, 1.0)
+    cam_quat = wp.quat_identity()

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 9d1059f8cc

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[Idate96]

Addressed CodeRabbit review nits in 71f669f: updated SPDX year to 2026, switched to wp.quat_identity() consistently, and added docstrings for the new regression test.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ All tests successful. No failed tests found.

📢 Thoughts on this report? Let us know!

[daniela-hase]

Thank you for catching and fixing this! :D

Looks all good to me!