The robotics company Unitree Technology recently announced that it will fully open-source its robot training source code, a move that has garnered widespread attention in the industry. The open-source content includes reinforcement learning (RL) training code, as well as complete code for both Sim-to-Sim and Sim-to-Real transitions.
This open-source project aims to assist developers and researchers in conducting robot simulation training in virtual environments, and to smoothly transfer the training results to physical robots, significantly shortening development cycles and reducing R&D costs. This means that developers no longer need to build training environments from scratch; they can directly use Unitree Technology's mature code to quickly achieve intelligent upgrades for robots.
It is reported that this technology has been successfully validated on multiple robots, including the Unitree H1, H1-2, and G1, with its stability and effectiveness fully proven. This open-source initiative will accelerate technological advancements in the robotics field and inject new vitality into industry innovation.
Project address: https://github.com/unitreerobotics/unitree_rl_gym