Reproduce “EigenDEXplore: Structured Exploration for Dexterous Manipulation with Human Priors” 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
Harsh Gupta, Tyler Ga Wei Lum, Changhao Wang, Chuer Pan, C. Karen Liu, Jeannette Bohg, Shuran Song
EigenDEXplore guides dexterous robot-hand exploration with coordinated perturbations derived from retargeted human hand motion, while retaining joint-space actions. The linear PCA basis encourages coordination patterns but does not constrain behavior. In four hardware tool-use tasks, policies trained in simulation reached an average 82% of demonstrated object-pose waypoints, versus 33% for Joint-Space; this measures task progress, not training time.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.