This repository enables training reinforcement learning policies in simulation environments (IsaacGym, IsaacSim, Gazebo, MuJoCo) and deploying them on physical robots. It supports multiple robot platforms including Unitree quadrupeds and humanoids, with ROS and ROS2 integration for real-world control. Neural network policies can be exported and run via libtorch or ONNXRuntime, and the framework includes pre-trained policies for locomotion and whole-body control tasks.