Easy to Use
GBC Project serves as an extension library for Isaac Lab, requiring only few external libraries and all of them are installable via pip. All sub projects for robots can be configured similar to standard Isaac Lab tasks, making it easy to use and integrate into existing workflows.
Retargeting for Any Humanoid
With the specially designed retargeting network and generalized scripts, GBC Project can create a fitted humanoid retargeting model for any architecture. Users can map any motion (following SMPL+H) to their own humanoid robot, with only a simple configuration setup and a few days of training!
Hybrid & Advanced RL
GBC Project supports advanced imitation learning combined with reinforcement learning, enabling both training with or without reference motions. The powerful backbone allows humanoids to imitate multiple complex motions, such as walking, running, jumping, dancing with a single policy file, and can function even on full-size humanoids like Unitree H1-2.
📰 Latest News
Stay updated with the latest developments and announcements