Reproduce “MoMa-NMP: Scene-Scale Neural Motion Planning for Mobile Manipulators” 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
Adam Kan, Jason Jingzhou Liu, Jiahui Yang, Kenneth Shaw, Deepak Pathak
The authors present MoMa-NMP, a learned whole-body planner for mobile manipulation from partial observations. It learns from global routes refined into safer action targets and receives on-policy correction using privileged scene geometry during training. Across 80 real-world trials in 10 unseen scenes, it succeeds in 62 (77.5%) versus 23/80 for cuRobo; classical planners plan once from initial geometry.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.