Language: English | 简体中文

MotrixLab

Train robot policies in simulation, then deploy them to real hardware.

Microduck locomotion policies trained with MotrixLab, rendered in MotrixRender — watch the HD video.

📖 Documentation: 简体中文 | English

Contents

• What is MotrixLab?
• Key Features
• Quick Start
• Task Environments
• Built-in Robot Models
• What's Inside
• Contributing
• Contact

What is MotrixLab?

MotrixLab is an open-source reinforcement learning framework for robot training, built on the high-performance MotrixSim physics engine. Define an environment once, train it with thousands of parallel environment instances using SKRL, RSL-RL, or the built-in FastSAC, and deploy the resulting policy to MuJoCo, native MotrixSim, or Unitree hardware — all through a single command-line interface.

Key Features

• Unified Interface: Provides a concise and unified reinforcement learning training and evaluation interface
• Multi-framework Support: Supports SKRL (JAX/PyTorch), RSLRL (PyTorch), and the built-in FastSAC implementation
• Rich Environments: Includes various robot simulation environments such as basic control, locomotion, and manipulation tasks
• Sim-to-Real Deployment: The same policy code deploys via the deploy CLI — simulation deployment to MuJoCo or native MotrixSim, hardware deployment to Unitree Go2
• High-precision, High-performance Simulation: Built on MotrixSim, a high-precision, high-performance physics engine
• Visual Training: Supports real-time rendering and training process visualization

🚀 Quick Start

Prerequisites

Requirement	Notes
Python 3.10.x	The workspace pins ==3.10.*
uv	Python project and dependency manager — installation guide
Git LFS	Robot meshes, motion data, and videos are tracked by LFS
OS	Linux x86_64 or Windows x86_64; the JAX training backend is Linux-only
GPU	NVIDIA (CUDA) or AMD (ROCm) — the matching wheels are selected automatically by sh install.sh

1. Clone the repository

git clone https://github.com/Motphys/MotrixLab
cd MotrixLab
git lfs pull

2. Install dependencies

Linux:

sh install.sh

Windows (PowerShell):

.\install.ps1
# if blocked by the execution policy:
powershell -ExecutionPolicy Bypass -File install.ps1

This auto-detects your GPU vendor (NVIDIA → CUDA, AMD → ROCm) and installs the runtime workspace packages with the matching PyTorch wheels, required SKRL Torch, and built-in FastSAC. Use --gpu cuda|rocm to override detection; --rslrl adds RSL-RL, while --skrl-jax adds the optional Linux-only SKRL JAX backend. --skrl-torch remains accepted as a compatibility flag because SKRL Torch is already required. Flags combine; see sh install.sh --help.

3. Train your first policy

Activate the installed environment (Windows PowerShell: .venv\Scripts\Activate.ps1):

source .venv/bin/activate
python scripts/train.py task=microduck-walk-flat/motrix.fastsac play=true

While training, the built-in dashboard shows live run progress, episode statistics, throughput, rewards, and system health:

Training runs thousands of parallel environment instances; when it finishes, the trained policy is loaded and played in the viewer automatically. Checkpoints and TensorBoard logs are saved under runs/microduck-walk-flat/; watch the curves with:

tensorboard --logdir runs/microduck-walk-flat

Training finishes in minutes: mean return and episode length typically converge after about 4,000 iterations:

4. Replay the trained policy

Replay the latest trained policy without retraining (for example, after stopping training early with Ctrl+C):

python scripts/play.py env=microduck-walk-flat

A trained microduck policy replayed in the viewer:

microduck-walk.mp4

🌍 Task Environments

MotrixLab ships 50+ built-in simulation environments spanning basic control, quadruped and humanoid locomotion, whole-body motion tracking, and manipulation. The main categories:

Preview	Category	Example environments
	Quadruped velocity tracking	go2-walk-flat · go2-walk-rough · go1-walk-rough · anymalc-walk-flat
	Humanoid velocity tracking	g1-walk-flat · k1-walk-rough · dex-evt-walk-flat · microduck-walk-flat
	Whole-body tracking (WBT)	g1-wbt-dance · k1-wbt-freekick · g1-29dof-wbt-largebox

python scripts/view.py env=go2-walk-rough

See the full environment gallery for all registered environments and their supported training algorithms.

🤖 Built-in Robot Models

Seven reusable robot models are registered out of the box and can be combined into any scene or task:

Screenshot	Registry name	Type	DoF
	anymal_c	Quadruped	12
	dex-evt	Humanoid	23
	g1-29dof	Humanoid	29
	go1	Quadruped	12
	go2	Quadruped	12
	k1	Humanoid	22
	microduck	Humanoid	14

python scripts/view.py robot=go2

Use the public configuration API to compose a model directly:

from motrix_robots import Microduck, UnitreeGo2Robot

microduck = Microduck()
go2 = UnitreeGo2Robot()

See Supported Robots for configuration details and how to add your own model.

🏗️ What's Inside

MotrixLab is a uv workspace with the following packages:

Package	PyPI name	Description
motrix_deploy	motrix-deploy	Artifacts, control sessions, runtime contracts, and deployment CLI
motrix_deploy_mujoco	motrix-deploy-mujoco	MuJoCo deployment backend plugin
motrix_deploy_motrixsim	motrix-deploy-motrixsim	Native MotrixSim deployment backend plugin
motrix_deploy_unitree	motrix-deploy-unitree	Asset-free Unitree hardware configuration, sensor/actuation wiring, and SDK2 DDS backend plugin
motrix_deploy_tasks	motrix-deploy-tasks	Walking tasks, deployment scenes, and installed Hydra recipes
motrix_env_core	motrix-env-core	Backend-agnostic environment framework, configuration, registry, and rendering
motrix_env_motrixsim	motrix-env-motrixsim	Live MotrixSim backend, renderer, and Torch frontend
motrix_env_mujoco	motrix-env-mujoco	Compile-only MuJoCo scene backend
motrix_robots	motrix-robots	Reusable robot configurations, default poses, and model assets
motrix_envs	motrix-envs	Built-in environments, task assets, data, and deployment-profile compilers
motrix_rl	motrix-rl	RL control plane, provider/trainer contracts, run/checkpoint handling, and discovery
motrix_rl_builtin	motrix-rl-builtin	Built-in Motrix FastSAC provider
motrix_rl_skrl	motrix-rl-skrl	Required SKRL PPO Torch provider; optional JAX extra
motrix_rl_rslrl	motrix-rl-rslrl	Optional RSL-RL PPO Torch provider

🤝 Contributing

Contributions are welcome! See CONTRIBUTING.md for the development environment setup, branch and commit conventions, and the configured checks (prek, ruff, dprint, mypy).

📬 Contact

Have questions or suggestions? Feel free to contact us through:

• GitHub Issues: Submit Issues
• Discussions: Join Discussion

Project policies

• Contributing
• Security policy
• Code of Conduct
• Support
• Third-party notices
• Changelog
• Citation metadata
• License