Counterexample-Guided Strategy Improvement for POMDPs Using Recurrent Neural Networks
This addresses the challenge of strategy synthesis in POMDPs for applications requiring provable guarantees, representing a strong specific gain rather than a broad paradigm shift.
The paper tackles the computationally intractable problem of synthesizing strategies for partially observable Markov decision processes (POMDPs) that adhere to probabilistic temporal logic constraints, by proposing a method combining recurrent neural networks (RNNs) and formal verification, resulting in up to three orders of magnitude improvement in solving times and model sizes.
We study strategy synthesis for partially observable Markov decision processes (POMDPs). The particular problem is to determine strategies that provably adhere to (probabilistic) temporal logic constraints. This problem is computationally intractable and theoretically hard. We propose a novel method that combines techniques from machine learning and formal verification. First, we train a recurrent neural network (RNN) to encode POMDP strategies. The RNN accounts for memory-based decisions without the need to expand the full belief space of a POMDP. Secondly, we restrict the RNN-based strategy to represent a finite-memory strategy and implement it on a specific POMDP. For the resulting finite Markov chain, efficient formal verification techniques provide provable guarantees against temporal logic specifications. If the specification is not satisfied, counterexamples supply diagnostic information. We use this information to improve the strategy by iteratively training the RNN. Numerical experiments show that the proposed method elevates the state of the art in POMDP solving by up to three orders of magnitude in terms of solving times and model sizes.