Papers.Data.Code
Repo

mjbatch

kevinzakka

Runs large MuJoCo batches on CPU and exposes live NumPy views for state, control, and model fields. Steps all sims in parallel from one Python call.

GitHub ↗Sep 10, 2026Paper ↗

  • rl
  • robotics
  • mujoco
  • batched-simulation
  • cpu-parallelism
Shows co-design vs stock comparison with quantitative improvement (+57%), the core claim
Shows co-design vs stock comparison with quantitative improvement (+57%), the core claim

What it is

Python library for batched MuJoCo on CPU, stepping thousands of simulations in parallel with direct NumPy access to state, controls, and per-sim model params. Run thousands of MuJoCo simulations in parallel on CPU from Python, with in-place array access to batched state, controls, and per-simulation model parameters.

Key results

  • Supports 4096 parallel sims in the README example
  • Uses a C++ thread pool and releases the Python GIL
  • Allows per-sim MjModel edits via expand() and set_const()

One useful ML email a week.

More from Sep 14 – Sep 20

The whole week →

Picked Sep 18

Picked Sep 17

Worth your time, not all of your time.

We track new ML work across papers, code, and data, then keep only the pieces worth a closer look. Selected, not collected.

Papers.Data.Code

The Weekly ML Signal

Selected, not collected.

One useful ML email a week: the papers, repos and datasets worth your attention.

Worth your time, not all of your time.