FetchMan: Vision-Based Humanoid Policy Trained in Simulation

kevin_zakka · x · 2026-08-22

Meet FetchMan: a vision-based humanoid policy trained entirely in simulation that achieves zero-shot transfer to diverse real-world scenes and objects. While simulation has produced impressive locomotion policies, this project explores how far the same recipe extends to vision-based loco-manipulation, demonstrating the potential of Sim2Real for complex tasks.

Original post →

More from Embodied

Embodied channel →