This library provides a collection of reinforcement learning environments for robotic manipulation and locomotion tasks, built upon NVIDIA's Isaac Lab. It targets researchers and developers in robotics and AI, offering a modular framework for rapid prototyping and experimentation with various robot morphologies and locomotion skills.

How It Works

The library leverages the Isaac Lab framework, utilizing its physics simulation capabilities and RL integration. Environments are designed to adhere to the OpenAI Gym API (v0.21.0), facilitating compatibility with standard RL training pipelines. Configurations for environments and agents are managed through Python classes or YAML files, allowing for flexible customization of tasks, robot embodiments, and observation/action spaces.

Quick Start & Requirements

• Installation: Clone the repository and install via pip install -e source/robot_lab using a Python interpreter with Isaac Lab installed.
• Prerequisites: Requires Isaac Lab v2.0.0 or higher.
• Verification: Run python scripts/tools/list_envs.py to list available environments.
• Documentation: Isaac Lab Documentation

Highlighted Details

• Supports a wide range of robot platforms including Unitree A1, Go2, B2, Anymal D, FFTAI GR1T1/GR1T2, Unitree H1/G1, Booster T1, RobotEra Xbot, and more.
• Includes implementations for various tasks such as velocity control, AMP (Adaptive Motion Primitives), and Handstand.
• Offers extensive command-line arguments for controlling training parameters, video recording, and resuming training.
• Provides tools for URDF to USD conversion for custom robot integration.

Maintenance & Community

• The project is actively maintained by Ziqi Fan.
• Discussion and support are available via GitHub Discussions and Discord.

Licensing & Compatibility

• The project's license is not explicitly stated in the README.
• Compatibility with commercial or closed-source projects is not specified.

Limitations & Caveats

• The README notes a potential issue with URDF conversion in the current Isaac Sim version, recommending Isaac Lab v1.4.1 and a specific conversion script for older versions.
• The project depends on Isaac Lab, which itself has significant hardware and software prerequisites.