Reproduce “GAE: General Action Expert for Real-Time Humanoid Teleoperation” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Yuefan Wang, Huaicheng Zhou, Xiao He, Zhijie He, Mingchuan Yang, Huayi Zhang, Li Chai, Jinxin Liu and 1 more
GAE learns to control humanoid avatars by separating feasible-motion generation from robust execution: a privileged simulation generator produces robot trajectories, while an executor learns from them under domain randomization and delayed human-motion inputs. With over 10,000 hours of motion data, GAE reaches 93.9% sequence success on the simulated Hard Set, versus 77.7% for SONIC, but does not perceive obstacles.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.