Reproduce “MIRROR: Reflecting Hidden Symmetries from Offline Robot Interaction Data” 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
Shiqi Liu, Yihang Yao, Peide Huang, Mengdi Xu, Ding Zhao
MIRROR learns paired transformations of robot observations and actions for a symmetry group supplied in advance. It trains on offline interaction data, using consistency constraints to align transformed transitions. In four simulated tasks, it reduced aggregate compatibility error to 0.717 from 1.838 for adapted LaLiGAN. Frame-averaged behavior cloning reached 85.12% mean success versus 80.48% for vanilla BC, showing results in these simulations.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.