Multi-cost bounded reachability in MDP

Abstract

We provide an efficient algorithm for multi-objective model-checking problems on Markov decision processes (MDPs) with multiple cost structures. The key problem at hand is to check whether there exists a scheduler for a given MDP such that all objectives over cost vectors are fulfilled. Reachability and expected cost objectives are covered and can be mixed. Empirical evaluation shows the algorithm’s scalability. We discuss the need for output beyond Pareto curves and exploit the available information from the algorithm to support decision makers.