Reproduce “LoComposition: Terrain-Adaptive Energy-Efficient Quadruped Locomotion without Gait 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
Loukas Kordos, Leonard T. Franz, Simon Rappenecker, Oliver Hausdoerfer, Angela P. Schoellig, Pavel Kolev, Georg Martius
LoComposition trains quadruped locomotion by separating command tracking, operational limits, energy use, and terrain perception. It penalizes mechanical power, uses stochastic terminations for soft-limit violations, and supplies a 13-by-11 elevation map. In Go2 simulation, LEP achieves COT of 0.28 ± 0.01 versus 0.64 ± 0.03 for RP; these results do not establish formal safety or hardware generalization.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.