Reproduce “LMGN: A Unified Policy Learning Model for Last-Mile Grasp-Navigation” 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
Xuenmin Hu, Qingxue Qian, Wei Luo, Benyi Ma, Bo Tang, Long Chen
LMGN guides quadruped robots from grasps to base stances and approach directions, coordinating base, arm, and gripper across task stages. It converts grasp poses into navigation cues and uses stage-specific policies. In Habitat, it reports 73.9% gaze success and 74.8% grasp success, above GAMMA; Isaac Gym ablations show removing action-level consistency yields the lowest mean among tested variants. Results are simulator-based.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.