Update pytorch kinematic solver benchmark param

[matafela]

Description

Update pytorch kinematic solver benchmark joint limits.

Type of change

• Enhancement (non-breaking change which improves an existing functionality)

Screenshots

Checklist

• I have run the black . command to format the code base.
• I have made corresponding changes to the documentation
• I have added tests that prove my fix is effective or that my feature works
• Dependencies have been updated, if applicable.

[Copilot]

Pull request overview

Updates the PyTorch kinematic solver benchmark configuration by changing the joint-limit ranges used for sampling and for initializing the solver.

Changes:

• Replaces the prior wide joint limits for the PyTorch benchmark with a narrow fixed range (1.0–2.5 rad for all joints), leaving the previous values commented out.
• Adds TODO/commented debug lines around investigating failed IK samples.

Comments suppressed due to low confidence (1)

scripts/benchmark/robotics/kinematic_solver/run_benchmark.py:435

• The commented-out failure debugging block (including an ipdb import) should be removed from the benchmark script. If this is useful, consider gating it behind a --debug flag and using structured logging rather than leaving commented debug code in the main path.

        cpu_t_err, cpu_r_err = get_pose_err(fk_xpos_cpu, check_xpos_cpu)

        cpu_result = {
            "cost_time_ms": cpu_elapsed * 1000.0,

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[Copilot]

Spelling/grammar in the TODO: "Easy to failed" should be corrected (e.g., "Easy to fail").

Suggested change

	# TODO: Easy to failed if use full joint range, consider adding a margin to avoid sampling near the joint limits.
	# TODO: Easy to fail if using the full joint range; consider adding a margin to avoid sampling near the joint limits.

[Copilot]

This TODO says to "consider adding a margin", but the benchmark already samples with a margin via _sample_qpos(...) (and passes margin=1e-1). As written it’s misleading—either remove it or update it to describe the actual failure mode and why the existing margin isn’t sufficient.

Suggested change

	# TODO: Easy to failed if use full joint range, consider adding a margin to avoid sampling near the joint limits.
	# NOTE: The benchmark already samples qpos with a nonzero margin in
	# `_sample_qpos(...)`. Even with that margin, benchmarking the Pytorch solver
	# over its full nominal joint range has proven unreliable in practice, so we
	# currently use a narrower range here to keep the benchmark stable.

[Copilot]

The new PYTORCH_*_LIMITS values (1.0–2.5 rad for every joint) both constrain sampling and are passed into PytorchSolverCfg.user_qpos_limits, changing the solver’s own joint-limit constraints. This likely makes the benchmark unrepresentative (and can artificially inflate success rate) compared to using the URDF limits with a sampling margin. Consider restoring the URDF/robot limits and only narrowing the sampling range (or compute a margin from the solver’s configured limits).

Suggested change

	# TODO: Easy to failed if use full joint range, consider adding a margin to avoid sampling near the joint limits.
	# PYTORCH_LOWER_LIMITS = [-6.2832, -6.2832, -3.1416, -6.2832, -6.2832, -6.2832]
	# PYTORCH_UPPER_LIMITS = [6.2832, 6.2832, 3.1416, 6.2832, 6.2832, 6.2832]
	PYTORCH_LOWER_LIMITS = [1.0, 1.0, 1.0, 1.0, 1.0, 1.0]
	PYTORCH_UPPER_LIMITS = [2.5, 2.5, 2.5, 2.5, 2.5, 2.5]
	# Keep the solver's configured joint limits aligned with the robot/URDF limits.
	# If benchmark sampling needs to avoid edge cases near joint boundaries, apply a
	# separate sampling margin at the sampling site rather than narrowing these
	# limits, because these values may also be passed into solver configuration.
	PYTORCH_LOWER_LIMITS = [-6.2832, -6.2832, -3.1416, -6.2832, -6.2832, -6.2832]
	PYTORCH_UPPER_LIMITS = [6.2832, 6.2832, 3.1416, 6.2832, 6.2832, 6.2832]

[Copilot]

PR description checklist indicates documentation and tests were updated/added, but this PR (as shown) only changes the benchmark script. If docs/tests were intended, they appear to be missing from the diff; otherwise, please update the description/checklist to match the actual changes.