Welcome to PredActor Discussions!

[MasterYip]

Welcome to the PredActor community!

PredActor is an end-to-end diffusion controller for humanoid robots. It jointly predicts future states and executable actions from proprioceptive feedback, supporting text-conditioned behaviors and joystick steering with 50 Hz onboard deployment on a Unitree G1.

This is a place to exchange ideas, ask questions, and share what you build with PredActor. We welcome discussions about:

• Getting started: installation, checkpoints, and running the demos.
• Research: diffusion control, state–action prediction, and motion guidance.
• Experiments: evaluation, reproduction, and adapting PredActor to new tasks or robots.
• Deployment: simulation, sim-to-real transfer, and onboard inference.
• Community projects: demos, extensions, and lessons learned.

When asking a technical question, please include your setup and relevant details so others can help. For reproducible bugs, please open a GitHub Issue with logs and steps to reproduce.

Resources

• Project page and demos
• Code
• Paper
• Models and artifacts
• Contact and community
• Discord

Feel free to introduce yourself below: what are you working on, and what would you like to explore with PredActor?