Isaacgym-Gazebo Sim2Sim for RL Quadruped Robot

Deploying quadruped robot policies trained in Isaac Gym to ROS2 Gazebo for Sim-to-Sim verification.

This repository provides an easy-to-use Sim2Sim pipeline for testing and validating reinforcement learning (RL) policies of quadruped robots within the ROS2 Gazebo environment.

Currently, three pretrained policies are supported:

[Walk-These-Ways]
Walk-These-Ways Demo

Margolis, Gabriel B., and Pulkit Agrawal.
"Walk these ways: Tuning robot control for generalization with multiplicity of behavior."
Conference on Robot Learning, PMLR, 2023.
👉 Project Website

[DreamWaQ]
DreamWaQ Demo

Nahrendra, I. Made Aswin, Byeongho Yu, and Hyun Myung.
"DreamWaQ: Learning robust quadrupedal locomotion with implicit terrain imagination via deep reinforcement learning."
ICRA 2023, IEEE.
👉 Project Website

[HIMLoco]
HIMLoco Demo

Long, J., Wang, Z., Li, Q., Cao, L., Gao, J., & Pang, J.
"Hybrid internal model: Learning agile legged locomotion with simulated robot response."
ICLR. 2024
👉 Project Website

Tested on ROS2 Humble and Ubuntu 22.04.

sudo apt install ros-humble-gazebo-ros2-control ros-humble-gazebo-ros2-control-demos
sudo apt install ros-humble-ros2-control ros-humble-controller-manager
sudo apt install ros-humble-gazebo-ros ros-humble-joint-state-publisher
sudo apt install ros-humble-gazebo-ros-pkgs

pip install torch==2.0.1+cu117 torchvision==0.15.2+cu117 torchaudio==2.0.2+cu117 \
 --index-url https://download.pytorch.org/whl/cu117

git clone https://github.com/evronix/quadruped_sim2sim.git
cd quadruped_sim2sim
colcon build --symlink-install
source install/setup.bash

ros2 launch robot_gazebo go1_gazebo.launch.py

• Walk-These-Ways

ros2 run robot_controller run_wtw_policy

• DreamWaQ

ros2 run robot_controller run_dreamwaq_policy

• HIMLoco

ros2 run robot_controller run_him_policy

ros2 run robot_UI run_ui

With the UI, you can control the robot's command_velocity interactively:

Coming soon...