This website requires JavaScript.

Policy-Guided Lazy Search with Feedback for Task and Motion Planning

Mohamed KhodeirAtharv SonwaneFlorian Shkurti
Oct 2022
摘要
PDDLStream solvers have recently emerged as viable solutions for Task andMotion Planning (TAMP) problems, extending PDDL to problems with continuousaction spaces. Prior work has shown how PDDLStream problems can be reduced to asequence of PDDL planning problems, which can then be solved usingoff-the-shelf planners. However, this approach can suffer from long runtimes.In this paper we propose LAZY, a solver for PDDLStream problems that maintainsa single integrated search over action skeletons, which gets progressively moregeometrically informed as samples of possible motions are lazily drawn duringmotion planning. We explore how learned models of goal-directed policies andcurrent motion sampling data can be incorporated in LAZY to adaptively guidethe task planner. We show that this leads to significant speed-ups in thesearch for a feasible solution evaluated over unseen test environments ofvarying numbers of objects, goals, and initial conditions. We evaluate our TAMPapproach by comparing to existing solvers for PDDLStream problems on a range ofsimulated 7DoF rearrangement/manipulation problems.
展开全部
图表提取

暂无人提供速读十问回答

论文十问由沈向洋博士提出,鼓励大家带着这十个问题去阅读论文,用有用的信息构建认知模型。写出自己的十问回答,还有机会在当前页面展示哦。

Q1论文试图解决什么问题?
Q2这是否是一个新的问题?
Q3这篇文章要验证一个什么科学假设?
0
被引用
笔记
问答