Robust Almost-Sure Reachability in Multi-Environment MDPs

Abstract Multiple-environment MDPs (MEMDPs) capture finite sets of that share the states but differ in transition dynamics. These models form a proper subclass partially observable (POMDPs). We consider synthesis policies robustly satisfy an almost-sure reachability property MEMDPs, is, one policy satisfies for all environments. For POMDPs, deciding existence robust is EXPTIME-complete problem. show this problem PSPACE-complete while require exponential memory general. exploit theoretical results to develop and implement algorithm shows promising synthesizing various benchmarks.