Reproduce “Distance Field Driven Reinforcement Learning for Reactive and Robust Collision Avoidance” 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
Taiga Ikeshima, Sheila Sutjipto, Lan Wu, Francisco J. Naranjo-Campos, Taiki Ishigaki, Taisuke Kobayashi, Teresa A. Vidal-Calleja, Eiichi Yoshida
The authors present a reinforcement-learning policy for robot avoidance using a Euclidean distance field and clearance gradient, with optional APF reward guidance. In simulation, the end-effector policy achieved 96% collision-free goal completion versus 86% for a simulation baseline, with faster successful trials but more collision events. On hardware, full-body policies succeeded in 7 of 10 fixed-obstacle trials; dynamic results were qualitative, and camera occlusions remained.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.