Berkeley Humanoid

What’s included

• Reinforcement learning
• Isaac Lab environments
• Sim-to-real transfer

Best for

Research into learned locomotion that carries over from simulation to real hardware.

How we work with humanoid robots

Humanoids are the most general-purpose robots, and the hardest to get right. We work with open humanoid designs and learning frameworks to prototype walking, balance and dexterous manipulation, training in simulation first and then transferring to hardware.

• Locomotion research with reinforcement learning
• Imitation learning for manipulation tasks
• Teleoperated data-collection rigs
• Custom humanoid prototypes from open CAD