[BP20b] Edoardo Bacci and David Parker. Probabilistic Guarantees for Safe Deep Reinforcement Learning. In Proc. 18th International Conference on Formal Modelling and Analysis of Timed Systems (FORMATS'20), Springer. To appear. September 2020. [pdf] [bib] [Proposes techniques for probabilistic verification of deep reinforcement learning policies, using PRISM as an underlying model checker.]
Downloads:  pdf pdf (617 KB)  bib bib
Notes: An extended version of the paper, with proofs, is available at https://arxiv.org/abs/2005.07073. The original publication is available at link.springer.com.
Abstract. Deep reinforcement learning has been successfully applied to many control tasks, but the application of such controllers in safety-critical scenarios has been limited due to safety concerns. Rigorous testing of these controllers is challenging, particularly when they operate in probabilistic environments due to, for example, hardware faults or noisy sensors. We propose MOSAIC, an algorithm for measuring the safety of deep reinforcement learning controllers in stochastic settings. Our approach is based on the iterative construction of a formal abstraction of a controller's execution in an environment, and leverages probabilistic model checking of Markov decision processes to produce probabilistic guarantees on safe behaviour over a finite time horizon. It produces bounds on the probability of safe operation of the controller for different initial configurations and identifies regions where correct behaviour can be guaranteed. We implement and evaluate our approach on controllers trained for several benchmark control problems.