google-deepmind/mujoco_playground

Health v2 · maintenance only

Adoption not part of the score

2250 stars · 364 forks observed · 2026-10-15

What it is AI-extracted, prompt v1, taxonomy v1, 2026-08-30, confidence not recorded

MuJoCo Playground is an open-source Python library of GPU-accelerated robot learning environments built on MuJoCo MJX and MuJoCo Warp. It streamlines simulation, reinforcement learning policy training, and sim-to-real transfer for quadrupeds, humanoids, dexterous hands, and robotic arms.

Use cases

• train locomotion policies for quadrupeds and humanoids on a single GPU
• run sim-to-real transfer of robot control policies
• train manipulation policies for dexterous hands and robotic arms
• learn vision-based robot policies from pixel inputs
• benchmark classic control environments like cartpole
• train PPO agents in GPU-parallel physics simulation

When to choose

• you need fast GPU-parallel RL training for robot control
• you want an integrated sim-to-real pipeline with MuJoCo MJX or Warp
• you need ready-made locomotion, manipulation, or vision-based environments
• you want to train policies in minutes on a single GPU

When to avoid

• you need CPU-only simulation without JAX or CUDA
• you need non-robotics general-purpose RL environments
• you require a stable long-term-support API for production systems
• you work outside Python or the JAX ecosystem

Facets

library · maturity active

machine-learning reinforcement-learning simulation robotics gpu-computing robotics machine-learning simulation gpu-computing python cross-platform mujoco mjx sim-to-real jax robot-learning locomotion manipulation physics-simulation gpu linux

Member repositories

For agents

markdown · JSON · MCP: product_card(name="google-deepmind/mujoco_playground")