Tracking Control for Petri Nets based on Monte-Carlo Tree Search

In this paper, an approach for tracking control of Petri nets based on Monte-Carlo Tree Search (MCTS) is proposed. The goal is to find a feasible firing sequence from an initial marking to a desired destination marking. The MCTS is a search algorithm based on random sampling of the search space. It has already gained interest in many game related applications. One of the best-known examples is AlphaGo. This paper shows the adaption of the MCTS algorithm to Petri nets and how MCTS can efficiently solve the tracking control problem. Additional methods for deadlock avoidance, supervisory control with forbidden markings and handling of uncontrollable transitions are shown. The proposed approach has a wide range of application like scheduling in flexible manufacturing systems or reachability analysis. An example illustrates the application of the MCTS tracking control to a flexible manufacturing system.