[release/3.0.0-beta2] Fix AutoMate PhysX sizing and assembly ID validation

[ooctipus]

Summary

• Cherry-pick Fix AutoMate PhysX sizing and assembly ID validation #6038 from develop into release/3.0.0-beta2
• Increase AutoMate PhysX GPU collision stack sizing for assembly/disassembly tasks
• Reject the literal ASSEMBLY_ID placeholder before resolving AutoMate asset paths

Validation

• git diff --check refs/remotes/upstream/release/3.0.0-beta2...HEAD
• python3 -m py_compile source/isaaclab_tasks/isaaclab_tasks/direct/automate/run_w_id.py source/isaaclab_tasks/isaaclab_tasks/direct/automate/run_disassembly_w_id.py source/isaaclab_tasks/isaaclab_tasks/direct/automate/assembly_env_cfg.py source/isaaclab_tasks/isaaclab_tasks/direct/automate/disassembly_env_cfg.py

[isaaclab-review-bot]

🤖 IsaacLab Review Bot — PR #6048

Summary

Cherry-pick of #6038 into release/3.0.0-beta2. Two fixes:

1. PhysX GPU collision stack sizing: Adds gpu_collision_stack_size=2**27 to both assembly and disassembly env configs
2. Assembly ID placeholder validation: Rejects the literal ASSEMBLY_ID placeholder string before launching simulation

✅ Findings

#	Category	Status	Notes
1	Correctness	✅	2**27 (128 MiB) collision stack is appropriate for complex contact-heavy assembly tasks at 128 envs
2	Consistency	✅	Both assembly_env_cfg.py and disassembly_env_cfg.py updated identically
3	Validation	✅	parser.error() is the correct argparse pattern — prints usage and exits with code 2
4	Error message	✅	Provides example assembly ID (00032) for user guidance
5	Changelog	✅	Fragment present at changelog.d/fix-automate-collision-stack.rst
6	CI	⏳	pre-commit ✅, Build Wheel ✅, installation tests pending

💡 Minor Observations (non-blocking)

• The validation only catches the exact string "ASSEMBLY_ID". If a user passes an empty string or whitespace, it would still proceed. Consider adding a regex check (e.g., r"^\d{5}$") for stricter validation in a future iteration.
• The PR body references #6038 but the title also mentions #6039 — the actual commit only cherry-picks from #6038. Title is slightly misleading but acceptable for a release branch backport.

Verdict: LGTM ✅

Clean, minimal cherry-pick. Both fixes are well-scoped and correctly applied. No concerns for merge.

[greptile-apps]

Greptile Summary

This cherry-pick from develop fixes two independent issues in the AutoMate simulation tasks: dropped PhysX contacts at the default 128-environment count, and accidental use of a documentation placeholder as an assembly ID.

• PhysX GPU collision stack (assembly_env_cfg.py, disassembly_env_cfg.py): adds gpu_collision_stack_size=2**27 (128 MB) alongside the existing gpu_max_rigid_contact_count and gpu_max_rigid_patch_count settings, keeping the assembly and disassembly configs in sync.
• Placeholder guard (run_w_id.py, run_disassembly_w_id.py): immediately after argument parsing, rejects the literal string "ASSEMBLY_ID" with a descriptive parser.error message before any filesystem or config-file side-effects occur.

Confidence Score: 4/5

Safe to merge; changes are small and targeted with no cross-cutting impact.

Both fixes are minimal and well-scoped. The gpu_collision_stack_size addition is a one-liner mirrored identically across the two env configs. The placeholder guard fires early and uses parser.error correctly. The one rough edge is in run_w_id.py: --assembly_id has no default and is not required, so omitting the flag entirely lets None reach update_task_param undetected by the new check.

run_w_id.py — the --assembly_id argument lacks a default and required=True, so the new placeholder check can be bypassed when the flag is omitted entirely.

Important Files Changed

Filename	Overview
source/isaaclab_tasks/isaaclab_tasks/direct/automate/assembly_env_cfg.py	Adds gpu_collision_stack_size=2**27 (128 MB) to PhysX GPU scene config to prevent dropped contacts at the default 128-environment count.
source/isaaclab_tasks/isaaclab_tasks/direct/automate/disassembly_env_cfg.py	Same gpu_collision_stack_size=2**27 addition as the assembly config, keeping the two environment configs in sync.
source/isaaclab_tasks/isaaclab_tasks/direct/automate/run_w_id.py	Adds a guard that rejects the literal ASSEMBLY_ID placeholder; --assembly_id has no default and no required=True, so a missing flag still bypasses the check and writes 'None' into the config.
source/isaaclab_tasks/isaaclab_tasks/direct/automate/run_disassembly_w_id.py	Same placeholder guard added; --assembly_id already has a concrete default (00731), so the check is only triggered by an explicit placeholder argument.
source/isaaclab_tasks/changelog.d/fix-automate-collision-stack.rst	New changelog entry accurately describing both the collision-stack size increase and the assembly-ID validation improvement.

Flowchart

%%{init: {'theme': 'neutral'}}%%
flowchart TD
    A([User invokes run script]) --> B[parse_args]
    B --> C{assembly_id == ASSEMBLY_ID?}
    C -- Yes --> D[parser.error: exit with message]
    C -- No --> E[update_task_param writes ID into cfg file]
    E --> F[subprocess runs isaaclab.sh train/play]
    F --> G([Simulation starts with PhysX gpu_collision_stack_size = 2 pow 27])

Loading

_{Reviews (1): Last reviewed commit: "Fix AutoMate PhysX collision stack sizin..." | Re-trigger Greptile}

[greptile-apps]

Missing None guard alongside the placeholder check

--assembly_id has no default value and is not marked required=True, so when a caller omits the flag entirely args.assembly_id is None. None != "ASSEMBLY_ID" is True, so the new validation passes silently and update_task_param receives None as the ID — causing the string literal 'None' to be written into the config file. Adding an is None check (or required=True on the argument) would close this gap alongside the placeholder guard.