Reproduce “Remember Where You Walk: Learning Spatial Memory for End-to-End Legged Robot Exploration” 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
Leon Magnus, Nico Bohlinger, Jan Peters, Oleg Arenz
LSMM gives a legged robot a place-indexed memory for exploring unknown mazes. Its estimated position selects a local window to update, while other locations retain their learned features. An auxiliary head trains the memory to predict maze walls. In simulation, reconstruction error fell from 0.33 to 0.10, more than three times lower than baselines, although coverage was comparable in tested mazes.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.