Reproduce “Learning to Move from and for Manipulation Policies: Spatializing Latent Action Representations for Goal-Pose-Free Mobile Manipulation” 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
Siyuan Zheng, Zhong Cao, Henry X. Liu
M4M estimates where a frozen manipulation policy can succeed and guides a mobile robot to a suitable stance before handing control to it. It learns from action-expert features and task outcomes, predicting candidate success from stereo views and instructions. Across eight simulated tasks, it achieves 47.0% equal-task macro success versus 26.5% for N2M; predictors are task-specific.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.