Fix OVPhysX contact force history reset

[AntoineRichard]

Summary

• clear filtered force-matrix history during OVPhysX contact-sensor resets
• preserve history for environments outside the reset mask
• add a CPU-only Warp regression test

Testing

• uv run --extra test python -m pytest source/isaaclab_ov/test/sensors/test_contact_sensor_kernels.py
• uv run python tools/changelog/cli.py check develop
• uv run isaaclab -f

[greptile-apps]

Greptile Summary

The PR fixes OVPhysX contact-sensor resets so filtered force-matrix history is cleared only for selected environments.

• Passes the force-matrix history buffer into the existing reset kernel.
• Clears every history and filter slot covered by the reset mask.
• Adds a CPU-only Warp regression test verifying both clearing and preservation behavior.
• Documents the fix in the OVPhysX changelog.

Confidence Score: 5/5

The PR appears safe to merge, with the reset behavior correctly scoped to selected environments.

The new loop uses dimensions consistent with the allocated history buffer, safely handles configurations without filtered-force buffers, and has focused regression coverage for masked resets.

Important Files Changed

Filename	Overview
source/isaaclab_ov/isaaclab_ov/sensors/contact_sensor/contact_sensor.py	Supplies the optional force-matrix history buffer to the reset kernel while preserving the existing environment mask.
source/isaaclab_ov/isaaclab_ov/sensors/contact_sensor/kernels.py	Clears all allocated filtered-force history entries for selected environments, with valid bounds and optional-buffer handling.
source/isaaclab_ov/test/sensors/test_contact_sensor_kernels.py	Adds focused CPU coverage proving selected history is cleared while unselected environment history remains unchanged.

_{Reviews (1): Last reviewed commit: "Fix OVPhysX contact force history reset" | Re-trigger Greptile}

[isaaclab-review-bot]

Isaac Lab Review Bot

The reset path now passes filtered force-matrix history into the existing Warp kernel, clears it only for masked environments, and includes a focused CPU regression test plus changelog fragment.

• Design and architecture: Using the existing reset kernel keeps history cleanup aligned with the other contact-sensor buffers and preserves unselected environments through the established env-mask early return.
• API: The positional kernel signature and the shown ContactSensor.reset launch site remain aligned. No public sensor property, configuration field, or exported symbol is changed; only reset semantics for force_matrix_w_history are corrected.
• Implementation: The added indexing matches the documented buffer layout, and the test verifies both clearing a selected environment and preserving an unselected one. The test directly exercises the kernel rather than the complete ContactSensor.reset integration path, leaving a small non-blocking integration-coverage tradeoff.

No blocking issues. No inline issue met the actionable-evidence threshold; the assessment above records the review feedback.

Automated review; human maintainers own approval decisions.

[camevor]

lgtm!