Point-Based Value Iteration Algorithms for Multi-Core Machines

Guy Shani

Recent scaling up of POMDP solvers towards realistic applications is largely due to point-based methods which quickly provide approximate solutions for medium-sized problems. New multi-core machines offer an opportunity to scale up to much larger domains. These machines support parallel execution and can speed up existing algorithms considerably. In this paper we suggest several ways in which point-based algorithms can be adapted to parallel computing. We overview the challenges and opportunities and present experimental evidence to the usability of our suggestions. Our results show that the opportunity lies mainly in parallelizing at the algorithmic level, not at the point-based backup level.

Subjects: 1.11 Planning; 12.1 Reinforcement Learning

Submitted: May 5, 2008

This page is copyrighted by AAAI. All rights reserved. Your use of this site constitutes acceptance of all of AAAI's terms and conditions and privacy policy.