A Bayesian network framework for stochastic discrete-event control

This article focuses on the use of Bayesian networks for stochastic discrete-event control applications. Bayesian networks offer several advantages for such applications, including a well-developed suite of efficient inference algorithms, model generality and compactness, and ease of model construction and/or model-learning. We show how we can formalise the control-theoretic semantics of a stochastic discrete-event control representation using a Bayesian network. We prove the space-efficiency of a Bayesian network relative to a probabilistic finite state machine. We demonstrate our approach on a simple elevator system