Schedule of Flexible Manufacturing Systems Based on Petri Nets and A* Search With a Neural Network Heuristic Function

An approach is presented to schedule a flexible manufacturing system (FMS) via Petri nets and A* algorithm with a heuristic function represented by a neural network. An algorithm is designed to calculate a minimal production time for each state of FMS by a place-timed Petri net. Consequently, the production data set is obtained by the algorithm, and is utilized to train a neural network such that it can play the role of a heuristic function to estimate a minimal time that it should be taken to complete all tasks given any state. Then, a schedule A* algorithm is developed via a place-timed Petri nets and neural network. Numerical experiments are carried out on a robotic arm handling system, and show that the propose approach is very efficient. Further, this work shows a way to capture heuristic rules for A* algorithm machine learning other than experts experiences.