Learning Local Heuristics for Search-Based Navigation Planning

Rishi VeerapaneniMuhammad Suhail SaleemMaxim Likhachev

Mar 2023

摘要

Graph search planning algorithms for navigation typically rely heavily onheuristics to efficiently plan paths. As a result, while such approachesrequire no training phase and can directly plan long horizon paths, they oftenrequire careful hand designing of informative heuristic functions. Recent workshave started bypassing hand designed heuristics by using machine learning tolearn heuristic functions that guide the search algorithm. While these methodscan learn complex heuristic functions from raw input, they i) require asignificant training phase and ii) do not generalize well to new maps andlonger horizon paths. Our contribution is showing that instead of learning aglobal heuristic estimate, we can define and learn local heuristics whichresults in a significantly smaller learning problem and improvesgeneralization. We show that using such local heuristics can reduce nodeexpansions by 2-20x while maintaining bounded suboptimality, are easy to train,and generalize to new maps & long horizon plans.

展开全部

图表提取

论文十问

参考文献

被引用

提问与回答

暂无人提供速读十问回答

被引用

笔记

问答

领域