Overview
legged_control2 is a C++ toolbox for legged robots. It provides basic state estimation and observer for debugging and analysis, and a unified hardware interface that facilitates the easy deployment of Reinforcement Learning (RL) policies on real-world robots.
This toolbox is implemented in C++, built on top of pinocchio and ros2_control, ensuring high-performance, real-time control for demanding robotic tasks.
legged_control2 supports multiple robots by reading different URDF provided by the user, and supports sim-to-sim (training simulation to validation simulation) with MuJoCo and Gazebo. Hardware interfaces shown below are supported for sim-to-real transfer.
unitree_ros2: Hardware interface for Unitree SDK1/2, such as the Unitree Go1 and G1.
booster_ros2: Hardware interface for Booster robotics.
soem_ros2: Basic interface for robots using EtherCAT Bus.
cleardrive_ros2: Hardware interface based on SOEM for robots using ClearDrive, such as the Berkeley Humanoid.
How to use the legged_control2?
legged_control2 can be easily installed on Ubuntu. The source code is also publicly available. To get started with the legged_control2, please refer to the Installation, and Getting Started pages.
Source code on GitHub: qiayuanl/legged_control2