Holosoma (Amazon FAR humanoid RL framework)

Holosoma（亚马逊人形 RL 训练部署框架）Advanced

An open-source humanoid reinforcement-learning framework for training and deployment, from Amazon's FAR team.

Holosoma is a humanoid robotics framework open-sourced on GitHub by Amazon's Frontier AI & Robotics (FAR) team. According to its repository, it connects “training a policy in simulation with reinforcement learning” to “deploying that policy on a real robot,” covering both velocity-command-based walking and whole-body motion tracking, able to train across multiple simulation backends and supporting humanoid platforms such as the Unitree G1. A common pain point in humanoid RL projects is that training code, the simulator, and real-robot deployment each end up as a separate, one-off setup, so switching robots or simulators means rebuilding almost everything; the value of a framework like this is in offering a ready-made pipeline. It sits in the same category as HumanoidVerse and legged_gym.

Related: Amazon Frontier AI & Robotics · HumanoidVerse · RL-based Locomotion Control · Sim-to-Real Transfer · Motion Tracking · Unitree G1
Sources: amazon-far/holosoma GitHub
As of: 2025