Guest Talk: Parameter-Independent Strategies for pMDPs

Monday, 19.11.2018, 3:00pm

Location: RWTH Aachen University, Department of Computer Science - Ahornstr. 55, building E3, room 9u10

Speaker: Sebastian Arming

 

Markov Decision Processes (MDPs) are a popular class of models suitable for solving control decision problems in probabilistic reactive systems. Parametric MDPs (pMDPs) include parameters in some of the transition probabilities to account for stochastic uncertainties of the environment such as noise or input disturbances. We study pMDPs with reachability objectives where the parameter values are unknown and impossible to measure directly during execution. We describe different types of strategies that are optimal for the whole parameter range, yielding different notions of „parameter-independent optimality“. We describe methods to compute such strategies. We evaluated our methods experimentally on several benchmarks: a motivating (repeated) learner model; a series of benchmarks of varying configurations of a robot moving on a grid; and a consensus protocol. This is joint work with Ezio Bartocci, Krishnendu Chatterjee, Joost-Pieter Katoen and Ana Sokolova.