Reproduce “SWAP: Test-Time Steerable Humanoid Whole-Body Collision Avoidance” 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
Zeji Yi, Guanqi He, Xiaofeng Guo, Guannan Qu, Pieter Abbeel, Yan Duan, Karen Liu, Guanya Shi
SWAP enables humanoids to propose and execute collision-aware whole-body motions. A perception-conditioned diffusion planner generates motion chunks, while a low-level reinforcement-learning controller tracks them and maintains balance. The planner learns from simulated replays of motions, and test-time steering selects among candidates. In simulation, steering raised sideways-walking success to 88.4% and lowered wall-cube contact to 2.2%, in tested settings.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.