Learning to Plan in High Dimensions via Neural Exploration-Exploitation Trees

Binghong Chen, Bo Dai, Qinjie Lin, Guo Ye, Han Liu, Le Song

Keywords: learning to plan, representation learning, learning to design algorithm, reinforcement learning, meta learning

TL;DR: We propose a meta path planning algorithm which exploits a novel attention-based neural module that can learn generalizable structures from prior experiences to drastically reduce the sample requirement for solving new path planning problems.

Abstract: We propose a meta path planning algorithm named \emph{Neural Exploration-Exploitation Trees~(NEXT)} for learning from prior experience for solving new path planning problems in high dimensional continuous state and action spaces. Compared to more classical sampling-based methods like RRT, our approach achieves much better sample efficiency in high-dimensions and can benefit from prior experience of planning in similar environments. More specifically, NEXT exploits a novel neural architecture which can learn promising search directions from problem structures. The learned prior is then integrated into a UCB-type algorithm to achieve an online balance between \emph{exploration} and \emph{exploitation} when solving a new problem. We conduct thorough experiments to show that NEXT accomplishes new planning problems with more compact search trees and significantly outperforms state-of-the-art methods on several benchmarks.

Code: https://github.com/NeurEXT/NEXT-learning-to-plan/blob/master/main.ipynb