Reproduce “PSR: Predictive Sensorimotor Representation Learning for Contact-Rich Manipulation” 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
Shengbao Li, Peng Xu, Chao Tang, Hao Wei, Jiaheng Wang, Hong Yin, Jiangtao Chen, Jinxuan Zhu and 3 more
PSR-VLA conditions a vision-language-action policy for contact-rich manipulation using predictive sensorimotor representations. It forecasts future contact signals and robot motion from images and histories of force/torque, joint states, and joint effort, then injects representations at multiple policy depths. Across six tasks, it succeeds in 110 of 120 trials (91.7%), versus 87 of 120 for an adapted ForceVLA2-π0.5 baseline; each method-task result uses one training run.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.