CNR ExploRA

Contributo in atti di convegno, 2016, ENG, 10.1007/978-3-319-43425-4_17

Policy learning for time-bounded reachability in continuous-time Markov decision processes via doubly-stochastic gradient ascent

Bartocci E.; Bortolussi L.; Brazdil T.; Milios D.; Sanguinetti G.

Faculty of Informatics, Vienna University of Technology, Vienna, Austria; University of Trieste, Trieste, Italy; CNR-ISTI, Pisa, Italy; Saarland University, Saarbrücken, Germany; Faculty of Informatics, Masaryk University, Brno, Czech Republic; School of Informatics, University of Edinburgh, Edinburgh, United Kingdom; Centre for Synthetic and Systems Biology, University of Edinburgh, Edinburgh, United Kingdom

Continuous-time Markov decision processes are an important class of models in a wide range of applications, ranging from cyberphysical systems to synthetic biology. A central problem is how to devise a policy to control the system in order to maximise the probability of satisfying a set of temporal logic specifications. Here we present a novel approach based on statistical model checking and an unbiased estimation of a functional gradient in the space of possible policies. The statistical approach has several advantages over conventional approaches based on uniformisation, as it can also be applied when the model is replaced by a black box, and does not suffer from state-space explosion. The use of a stochastic gradient to guide our search considerably improves the efficiency of learning policies. We demonstrate the method on a proof-ofprinciple non-linear population model, showing strong performance in a non-trivial task.

13th International Conference on Quantitative Evaluation of Systems, QEST 2016, pp. 244–259, Quebec City, Canada, 23-25 August, 2016