# google-deepmind/mujoco_playground architecture

> Architecture diagram of the GitHub repository google-deepmind/mujoco_playground, generated by GitDiagram from its file tree, README and a sample of source files (last updated 2026-11-19). It is an AI-made overview: check details against the source.

- Interactive diagram: https://gitdiagram.com/google-deepmind/mujoco_playground
- Repository: https://github.com/google-deepmind/mujoco_playground

## Overview

MuJoCo Playground provides robot simulation environments, RL training and playback, and workflows connecting simulation with physical hardware. The principal path is selecting a registered environment, training or playing a policy, and optionally deploying it; separate paths support teleoperation, system identification, and agentic manipulation. The README documents these capabilities, while source-index references and excerpts support selected internal links. Several task and hardware implementations were not sampled, so their internal call relationships are not inferred. The older RSS 2025 JAX/MJX environments remain a distinct legacy branch.

Main entry points: [`mujoco_playground/_src/rl/env_registry.py`](https://github.com/google-deepmind/mujoco_playground/blob/main/mujoco_playground/_src/rl/env_registry.py) (Environment registry), [`mujoco_playground/rl/rsl_rl/train.py`](https://github.com/google-deepmind/mujoco_playground/blob/main/mujoco_playground/rl/rsl_rl/train.py) (Policy training), [`mujoco_playground/rl/rsl_rl/play.py`](https://github.com/google-deepmind/mujoco_playground/blob/main/mujoco_playground/rl/rsl_rl/play.py) (Policy playback), [`rss2025/mujoco_playground/_src/mjx_env.py`](https://github.com/google-deepmind/mujoco_playground/blob/main/rss2025/mujoco_playground/_src/mjx_env.py) (MJX environments)

## Diagram (Mermaid)

```mermaid
flowchart TD

subgraph group_simulation["Simulation and tasks"]
  node_robot_models["Robot models"]
  node_task_suite["Simulation tasks"]
  node_pick_cube["Pick cube<br/>[task.py]"]
  node_whole_body["Whole-body tracking<br/>[task.py]"]
  node_agentic["Agentic tabletop"]
  node_ik["Cartesian IK<br/>[ik_controller.py]"]
end

subgraph group_learning["RL learning"]
  node_registry["Environment registry<br/>[env_registry.py]"]
  node_train["Policy training<br/>[train.py]"]
  node_play["Policy playback<br/>[play.py]"]
  node_vector_env["Vector environment adapter<br/>[vecenv_wrapper.py]"]
  node_checkpoint["Policy checkpoints<br/>[checkpoint.py]"]
end

subgraph group_hardware["Hardware workflows"]
  node_teleop["SO-101 teleoperation"]
  node_sysid["System identification"]
  node_deployment["Pick-cube deployment"]
  node_robot_hardware["Robot hardware drivers<br/>[device.py]"]
end

subgraph group_api["Runtime foundation"]
  node_wrappers["Environment wrappers<br/>[wrapper.py]"]
  node_device_wrappers["Device-layer adapters"]
  node_assets["Asset resolution<br/>[paths.py]"]
  node_rng["Random seeding<br/>[rng.py]"]
  node_compute["Compute selection<br/>[xpu.py]"]
end

subgraph group_legacy["RSS 2025 legacy"]
  node_legacy_env["MJX environments<br/>[mjx_env.py]"]
  node_legacy_locomotion["Locomotion tasks"]
  node_legacy_manipulation["Manipulation tasks"]
  node_legacy_learning["Legacy training<br/>[train_jax_ppo.py]"]
end

node_engineer(("Robot developer"))

node_engineer -->|"selects environment"| node_registry
node_registry -->|"registers task"| node_pick_cube
node_registry -->|"registers task"| node_whole_body
node_registry -->|"builds environments"| node_task_suite
node_train -->|"loads environment"| node_registry
node_train -->|"wraps environment"| node_vector_env
node_vector_env -->|"seeds episodes"| node_rng
node_train -->|"selects device"| node_compute
node_play -->|"loads environment"| node_registry
node_play -->|"wraps environment"| node_vector_env
node_play -->|"loads policy"| node_checkpoint
node_engineer -->|"trains policy"| node_train
node_engineer -->|"plays policy"| node_play
node_engineer -->|"controls robot"| node_teleop
node_teleop -->|"controls follower"| node_robot_hardware
node_engineer -->|"fits dynamics"| node_sysid
node_sysid -->|"uses hardware logs"| node_robot_hardware
node_engineer -->|"deploys policy"| node_deployment
node_deployment -->|"runs on hardware"| node_robot_hardware
node_engineer -->|"starts task"| node_agentic
node_agentic -->|"uses IK controller"| node_ik
node_wrappers -->|"exports adapters"| node_device_wrappers
node_engineer -->|"trains legacy tasks"| node_legacy_learning
node_legacy_learning -.->|"runs MJX tasks"| node_legacy_env
node_legacy_env -.->|"supports tasks"| node_legacy_locomotion
node_legacy_env -.->|"supports tasks"| node_legacy_manipulation

click node_robot_models "https://github.com/google-deepmind/mujoco_playground/tree/main/mujoco_playground/robots"
click node_task_suite "https://github.com/google-deepmind/mujoco_playground/tree/main/mujoco_playground/tasks"
click node_pick_cube "https://github.com/google-deepmind/mujoco_playground/blob/main/mujoco_playground/tasks/manipulation/pick_cube/so101/task.py"
click node_whole_body "https://github.com/google-deepmind/mujoco_playground/blob/main/mujoco_playground/tasks/whole_body/tracking/task.py"
click node_agentic "https://github.com/google-deepmind/mujoco_playground/tree/main/mujoco_playground/tasks/agentic/tabletop_v0"
click node_ik "https://github.com/google-deepmind/mujoco_playground/blob/main/mujoco_playground/tasks/agentic/tabletop_v0/ik_controller.py"
click node_registry "https://github.com/google-deepmind/mujoco_playground/blob/main/mujoco_playground/_src/rl/env_registry.py"
click node_train "https://github.com/google-deepmind/mujoco_playground/blob/main/mujoco_playground/rl/rsl_rl/train.py"
click node_play "https://github.com/google-deepmind/mujoco_playground/blob/main/mujoco_playground/rl/rsl_rl/play.py"
click node_vector_env "https://github.com/google-deepmind/mujoco_playground/blob/main/mujoco_playground/_src/rl/rsl_rl/vecenv_wrapper.py"
click node_checkpoint "https://github.com/google-deepmind/mujoco_playground/blob/main/mujoco_playground/_src/rl/rsl_rl/checkpoint.py"
click node_teleop "https://github.com/google-deepmind/mujoco_playground/tree/main/mujoco_playground/tasks/teleop/so101"
click node_sysid "https://github.com/google-deepmind/mujoco_playground/tree/main/mujoco_playground/tasks/sysid"
click node_deployment "https://github.com/google-deepmind/mujoco_playground/tree/main/mujoco_playground/tasks/manipulation/pick_cube/so101/deploy"
click node_robot_hardware "https://github.com/google-deepmind/mujoco_playground/blob/main/mujoco_playground/robots/so101/real/device.py"
click node_wrappers "https://github.com/google-deepmind/mujoco_playground/blob/main/mujoco_playground/_src/wrapper/wrapper.py"
click node_device_wrappers "https://github.com/google-deepmind/mujoco_playground/blob/main/mujoco_playground/_src/wrapper/device_layer_wrapper.py"
click node_assets "https://github.com/google-deepmind/mujoco_playground/blob/main/mujoco_playground/_src/paths.py"
click node_rng "https://github.com/google-deepmind/mujoco_playground/blob/main/mujoco_playground/_src/rl/rng.py"
click node_compute "https://github.com/google-deepmind/mujoco_playground/blob/main/mujoco_playground/_src/xpu.py"
click node_legacy_env "https://github.com/google-deepmind/mujoco_playground/blob/main/rss2025/mujoco_playground/_src/mjx_env.py"
click node_legacy_locomotion "https://github.com/google-deepmind/mujoco_playground/tree/main/rss2025/mujoco_playground/_src/locomotion"
click node_legacy_manipulation "https://github.com/google-deepmind/mujoco_playground/tree/main/rss2025/mujoco_playground/_src/manipulation"
click node_legacy_learning "https://github.com/google-deepmind/mujoco_playground/blob/main/rss2025/learning/train_jax_ppo.py"

classDef toneNeutral fill:#f8fafc,stroke:#334155,stroke-width:1.5px,color:#0f172a
classDef toneBlue fill:#dbeafe,stroke:#2563eb,stroke-width:1.5px,color:#172554
classDef toneAmber fill:#fef3c7,stroke:#d97706,stroke-width:1.5px,color:#78350f
classDef toneMint fill:#dcfce7,stroke:#16a34a,stroke-width:1.5px,color:#14532d
classDef toneRose fill:#ffe4e6,stroke:#e11d48,stroke-width:1.5px,color:#881337
classDef toneIndigo fill:#e0e7ff,stroke:#4f46e5,stroke-width:1.5px,color:#312e81
classDef toneTeal fill:#ccfbf1,stroke:#0f766e,stroke-width:1.5px,color:#134e4a
class node_robot_models,node_task_suite,node_pick_cube,node_whole_body,node_agentic,node_ik toneBlue
class node_registry,node_train,node_play,node_vector_env,node_checkpoint toneAmber
class node_teleop,node_sysid,node_deployment,node_robot_hardware toneMint
class node_wrappers,node_device_wrappers,node_assets,node_rng,node_compute toneRose
class node_legacy_env,node_legacy_locomotion,node_legacy_manipulation,node_legacy_learning,node_engineer toneIndigo
```

## Components

### Simulation and tasks

- [Robot models](https://github.com/google-deepmind/mujoco_playground/tree/main/mujoco_playground/robots): `mujoco_playground/robots`
- [Simulation tasks](https://github.com/google-deepmind/mujoco_playground/tree/main/mujoco_playground/tasks): `mujoco_playground/tasks`
- [Pick cube](https://github.com/google-deepmind/mujoco_playground/blob/main/mujoco_playground/tasks/manipulation/pick_cube/so101/task.py): `mujoco_playground/tasks/manipulation/pick_cube/so101/task.py`
- [Whole-body tracking](https://github.com/google-deepmind/mujoco_playground/blob/main/mujoco_playground/tasks/whole_body/tracking/task.py): `mujoco_playground/tasks/whole_body/tracking/task.py`
- [Agentic tabletop](https://github.com/google-deepmind/mujoco_playground/tree/main/mujoco_playground/tasks/agentic/tabletop_v0): `mujoco_playground/tasks/agentic/tabletop_v0`
- [Cartesian IK](https://github.com/google-deepmind/mujoco_playground/blob/main/mujoco_playground/tasks/agentic/tabletop_v0/ik_controller.py): `mujoco_playground/tasks/agentic/tabletop_v0/ik_controller.py`

### RL learning

- [Environment registry](https://github.com/google-deepmind/mujoco_playground/blob/main/mujoco_playground/_src/rl/env_registry.py): `mujoco_playground/_src/rl/env_registry.py`
- [Policy training](https://github.com/google-deepmind/mujoco_playground/blob/main/mujoco_playground/rl/rsl_rl/train.py): `mujoco_playground/rl/rsl_rl/train.py`
- [Policy playback](https://github.com/google-deepmind/mujoco_playground/blob/main/mujoco_playground/rl/rsl_rl/play.py): `mujoco_playground/rl/rsl_rl/play.py`
- [Vector environment adapter](https://github.com/google-deepmind/mujoco_playground/blob/main/mujoco_playground/_src/rl/rsl_rl/vecenv_wrapper.py): `mujoco_playground/_src/rl/rsl_rl/vecenv_wrapper.py`
- [Policy checkpoints](https://github.com/google-deepmind/mujoco_playground/blob/main/mujoco_playground/_src/rl/rsl_rl/checkpoint.py): `mujoco_playground/_src/rl/rsl_rl/checkpoint.py`

### Hardware workflows

- [SO-101 teleoperation](https://github.com/google-deepmind/mujoco_playground/tree/main/mujoco_playground/tasks/teleop/so101): `mujoco_playground/tasks/teleop/so101`
- [System identification](https://github.com/google-deepmind/mujoco_playground/tree/main/mujoco_playground/tasks/sysid): `mujoco_playground/tasks/sysid`
- [Pick-cube deployment](https://github.com/google-deepmind/mujoco_playground/tree/main/mujoco_playground/tasks/manipulation/pick_cube/so101/deploy): `mujoco_playground/tasks/manipulation/pick_cube/so101/deploy`
- [Robot hardware drivers](https://github.com/google-deepmind/mujoco_playground/blob/main/mujoco_playground/robots/so101/real/device.py): `mujoco_playground/robots/so101/real/device.py`

### Runtime foundation

- [Environment wrappers](https://github.com/google-deepmind/mujoco_playground/blob/main/mujoco_playground/_src/wrapper/wrapper.py): `mujoco_playground/_src/wrapper/wrapper.py`
- [Device-layer adapters](https://github.com/google-deepmind/mujoco_playground/blob/main/mujoco_playground/_src/wrapper/device_layer_wrapper.py): `mujoco_playground/_src/wrapper/device_layer_wrapper.py`
- [Asset resolution](https://github.com/google-deepmind/mujoco_playground/blob/main/mujoco_playground/_src/paths.py): `mujoco_playground/_src/paths.py`
- [Random seeding](https://github.com/google-deepmind/mujoco_playground/blob/main/mujoco_playground/_src/rl/rng.py): `mujoco_playground/_src/rl/rng.py`
- [Compute selection](https://github.com/google-deepmind/mujoco_playground/blob/main/mujoco_playground/_src/xpu.py): `mujoco_playground/_src/xpu.py`

### RSS 2025 legacy

- [MJX environments](https://github.com/google-deepmind/mujoco_playground/blob/main/rss2025/mujoco_playground/_src/mjx_env.py): `rss2025/mujoco_playground/_src/mjx_env.py`
- [Locomotion tasks](https://github.com/google-deepmind/mujoco_playground/tree/main/rss2025/mujoco_playground/_src/locomotion): `rss2025/mujoco_playground/_src/locomotion`
- [Manipulation tasks](https://github.com/google-deepmind/mujoco_playground/tree/main/rss2025/mujoco_playground/_src/manipulation): `rss2025/mujoco_playground/_src/manipulation`
- [Legacy training](https://github.com/google-deepmind/mujoco_playground/blob/main/rss2025/learning/train_jax_ppo.py): `rss2025/learning/train_jax_ppo.py`

### Other components

- **Robot developer**

## Connections

- Robot developer → Environment registry: selects environment (evidence: `README.md`)
- Environment registry → Pick cube: registers task (evidence: `mujoco_playground/_src/rl/env_registry.py`)
- Environment registry → Whole-body tracking: registers task (evidence: `mujoco_playground/_src/rl/env_registry.py`)
- Environment registry → Simulation tasks: builds environments (evidence: `mujoco_playground/_src/rl/env_registry.py`)
- Policy training → Environment registry: loads environment (evidence: `mujoco_playground/rl/rsl_rl/train.py`)
- Policy training → Vector environment adapter: wraps environment (evidence: `mujoco_playground/rl/rsl_rl/train.py`)
- Vector environment adapter → Random seeding: seeds episodes (evidence: `mujoco_playground/_src/rl/rsl_rl/vecenv_wrapper.py`)
- Policy training → Compute selection: selects device (evidence: `mujoco_playground/rl/rsl_rl/train.py`)
- Policy playback → Environment registry: loads environment (evidence: `mujoco_playground/rl/rsl_rl/play.py`)
- Policy playback → Vector environment adapter: wraps environment (evidence: `mujoco_playground/rl/rsl_rl/play.py`)
- Policy playback → Policy checkpoints: loads policy (evidence: `mujoco_playground/rl/rsl_rl/play.py`)
- Robot developer → Policy training: trains policy (evidence: `README.md`)
- Robot developer → Policy playback: plays policy (evidence: `README.md`)
- Robot developer → SO-101 teleoperation: controls robot (evidence: `README.md`)
- SO-101 teleoperation → Robot hardware drivers: controls follower (evidence: `README.md`)
- Robot developer → System identification: fits dynamics (evidence: `README.md`)
- System identification → Robot hardware drivers: uses hardware logs (evidence: `README.md`)
- Robot developer → Pick-cube deployment: deploys policy (evidence: `README.md`)
- Pick-cube deployment → Robot hardware drivers: runs on hardware (evidence: `README.md`)
- Robot developer → Agentic tabletop: starts task (evidence: `README.md`)
- Agentic tabletop → Cartesian IK: uses IK controller (evidence: `mujoco_playground/tasks/agentic/tabletop_v0/ik_controller.py`)
- Environment wrappers → Device-layer adapters: exports adapters (evidence: `mujoco_playground/wrapper.py`)
- Robot developer → Legacy training: trains legacy tasks (evidence: `README.md`)
- Legacy training → MJX environments: runs MJX tasks
- MJX environments → Locomotion tasks: supports tasks (evidence: `rss2025/mujoco_playground/_src/locomotion/g1/joystick.py`)
- MJX environments → Manipulation tasks: supports tasks
