Skip to content

Benchmark suite overhaul: consolidate aloha, add clutter & unitree G1#1325

Merged
erikfrey merged 6 commits into
google-deepmind:mainfrom
erikfrey:benchmark-suite-overhaul
May 4, 2026
Merged

Benchmark suite overhaul: consolidate aloha, add clutter & unitree G1#1325
erikfrey merged 6 commits into
google-deepmind:mainfrom
erikfrey:benchmark-suite-overhaul

Conversation

@erikfrey

@erikfrey erikfrey commented May 1, 2026

Copy link
Copy Markdown
Collaborator

Summary

Overhaul of the benchmark suite with structural consolidation, new benchmarks, and infrastructure improvements.

Benchmark changes

  • Consolidate aloha benchmarks: Merge aloha_pot/, aloha_sdf/, aloha_cloth/ into a single benchmarks/aloha/ directory with scene-specific XMLs (scene_pot.xml, scene_sdf.xml, scene_cloth.xml)
  • Add aloha_clutter: New benchmark with YCB/GSO objects from aloha_sim, exercising broad-phase collision detection with many convex mesh pairs
  • Replace apptronik_apollo with unitree_g1: New robot model with flat and hfield scenes and a shuffle dance replay trajectory
  • Delete config.txt: No longer needed

Infrastructure

  • Migrate run.py to pathlib: Replace os.path / glob with pathlib.Path throughout
  • Glob asset support: Asset specs in __init__.py now support * patterns (e.g. "ycb/*/google_64k") for flexible directory mapping
  • Auto nstep from replay: When --replay is specified and --nstep is not, nstep defaults to the trajectory length instead of 1000

Documentation

  • Update benchmarks/README.md: Fix output format example, document glob syntax, correct SSH→HTTPS URLs

Merge aloha_pot, aloha_sdf, aloha_cloth into benchmarks/aloha/ with
scene-specific XMLs (scene_pot.xml, scene_sdf.xml, scene_cloth.xml).

Add aloha_clutter benchmark with YCB/GSO object assets from aloha_sim.

Other changes:
- Add glob pattern support in run.py asset specs for flexible asset mapping
- Migrate run.py from os.path to pathlib
- Auto-set nstep from replay trajectory length when not explicitly specified
- Update benchmarks/README.md documentation
- Generalize load_trajectory docstring and variable naming
@erikfrey erikfrey force-pushed the benchmark-suite-overhaul branch from cb77c7c to 3ac052a Compare May 1, 2026 22:20
@erikfrey erikfrey changed the title Benchmark suite overhaul: consolidate aloha, add clutter & unitree G1, modernize run.py Benchmark suite overhaul: consolidate aloha, add clutter & unitree G1 May 1, 2026
@erikfrey erikfrey requested a review from thowell May 1, 2026 22:22
Comment thread benchmarks/aloha/README.md Outdated
Comment thread benchmarks/aloha/README.md
Comment thread benchmarks/aloha/scene_clutter.xml Outdated

Copy link
Copy Markdown
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

this file appears twice, is that intentional? here (unitree_g1/assets/) and in unitree_g1/

Copy link
Copy Markdown
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

do we know where this asset is from?

Copy link
Copy Markdown
Collaborator Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I think it's from a .to_zip() of the scene as constructed via mjspec in mjlab. We reference mjlab explicitly in the benchmark documentation so I think we're good here.

Comment thread benchmarks/unitree_g1/README.md Outdated
Comment thread benchmarks/unitree_g1/unitree_g1_mjlab.xml
@erikfrey erikfrey merged commit b655026 into google-deepmind:main May 4, 2026
10 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants