TY - JOUR AU - Sunberg, Zachary AU - Kochenderfer, Mykel PY - 2018/06/15 Y2 - 2024/03/29 TI - Online Algorithms for POMDPs with Continuous State, Action, and Observation Spaces JF - Proceedings of the International Conference on Automated Planning and Scheduling JA - ICAPS VL - 28 IS - 1 SE - Main Track DO - 10.1609/icaps.v28i1.13882 UR - https://ojs.aaai.org/index.php/ICAPS/article/view/13882 SP - 259-263 AB - <p> Online solvers for partially observable Markov decision processes have been applied to problems with large discrete state spaces, but continuous state, action, and observation spaces remain a challenge. This paper begins by investigating double progressive widening (DPW) as a solution to this challenge. However, we prove that this modification alone is not sufficient because the belief representations in the search tree collapse to a single particle causing the algorithm to converge to a policy that is suboptimal regardless of the computation time. This paper proposes and evaluates two new algorithms, POMCPOW and PFT-DPW, that overcome this deficiency by using weighted particle filtering. Simulation results show that these modifications allow the algorithms to be successful where previous approaches fail. </p> ER -