Dynamically constructed (PO)MDPs for adaptive robot planning

摘要: To operate in human-robot coexisting environments, intelligent robots need to simultaneously reason with commonsense knowledge and plan under uncertainty. Markov decision processes (MDPs) partially observable MDPs (POMDPs), are good at planning uncertainty toward maximizing long-term rewards; P-LOG, a declarative programming language Answer Set semantics, is strong common-sense reasoning. In this paper, we present novel algorithm called iCORPP dynamically about, construct (PO)MDPs using P-LOG. successfully shields exogenous domain attributes from (PO)MDPs, which limits computational complexity enables adapt the value changes these produce. We conduct number of experimental trials two example problems simulation demonstrate on real robot. Results show significant improvements compared competitive baselines.