# robfiras/loco-mujoco

Imitation learning benchmark focusing on complex locomotion tasks using MuJoCo.

Repository: https://github.com/robfiras/loco-mujoco
Canonical: https://ross.abutalabs.com/products/loco-mujoco
Homepage: https://loco-mujoco.readthedocs.io/
Language: Python
License: MIT
License Family: permissive
Last push: 2026-08-22T19:06:32+00:00

## Health v2 (maintenance only)
Score: 76/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 99, release rhythm 42, longevity 83
- inputs: {"age_days": 1168, "days_push": 11, "days_rel": 176, "gap_med": 271.0, "n_releases_24m": 3}
- flags: none
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 1453, forks 154 (observed 2026-08-28T04:04:46.647968+00:00)

## What it is
LocoMuJoCo is an imitation learning benchmark for whole-body locomotion control built on MuJoCo, featuring humanoid, quadruped, and musculoskeletal environments with retargeted motion capture datasets. It also supports reinforcement learning via custom rewards and includes JIT-compiled JAX reference algorithms like PPO, GAIL, AMP, and DeepMimic.

## Use cases
- benchmark imitation learning algorithms for locomotion
- train humanoid locomotion policies from motion capture data
- run GAIL or AMP on quadruped environments
- retarget motion capture datasets between robots
- compare trajectories with dynamic time warping metrics
- train RL agents with domain and terrain randomization

## When to choose
- you need a standardized benchmark for whole-body imitation learning
- you want motion-capture datasets pre-retargeted to humanoid or quadruped models
- you need fast GPU-parallel simulation via MJX/MjWarp
- you want both RL and IL support in one locomotion suite

## When to avoid
- you need manipulation or non-locomotion tasks
- you require a non-Python or non-JAX workflow
- you need real robot deployment out of the box

## Facets
- artifact type: library
- maturity: active
- function: machine-learning, reinforcement-learning, simulation, benchmarking
- domain: machine-learning, robotics, simulation, artificial-intelligence
- platform: python
- tags: imitation-learning, mujoco, locomotion, mocap, jax, gymnasium, biomechanics, humanoids, quadrupeds, dataset, linux, macos, gpu

## Member repositories
- robfiras/loco-mujoco (main) score 76

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:04:46.647968+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T04:35:45.822723+00:00, confidence not recorded.
  - readme: https://github.com/robfiras/loco-mujoco (fetched 2026-08-28T04:04:46.647968+00:00, sha 5d0f5ff14690)
  - registry_pypi: https://pypi.org/pypi/loco-mujoco/json (fetched 2026-08-29T11:45:31.377025+00:00, sha a5fb62e0b849)
- Data as of 2026-08-30T08:39:29.467469+00:00.