Point-Based Value Iteration Algorithms for Multi-Core Machines

Recent scaling up of POMDP solvers towards realistic appli-cations 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 exe-cution and can speed up existing algorithms considerably. In this paper we suggest several ways in which point-based al-gorithms can be adapted to parallel computing. We overview the challenges and opportunities and present experimental ev-idence to the usability of our suggestions. Our results show that the opportunity lies mainly in parallelizing at the algo-rithmic level, not at the point-based backup level