Deadlock Avoidance in Automated Manufacturing Systems Using Finite Automata and State Space Search

Abstract

An approach to deadlock avoidance based on finite automata is reviewed in this chapter. This approach begins from the framework introduced by Ramadge and Wonham (R&W) for modeling and control of discrete event systems based on formal languages generated by finite automata. We apply this framework to the problem of dynamic scheduling and control of automated manufacturing systems. A typical automated manufacturing system is composed of multiple machines and workstations that perform various operations on a part, and a material handling system that interconnects these machines and workstations. Parts are processed to completion by routing them through various machines and workstations according to their individual process plans. After processing is complete, the part leaves the system. Deadlock occurs when parts enter a "circular wait" condition where, in order to continue processing, a set of two or more parts require resources that are held by parts of the same set. Our approach to avoiding deadlock is unique in the following contributions: 1) A simple and natural way of formulating the "requirements model" of the R&W framework from the part routing plans, 2) An ability to handle parts with multiple routing plans within the framework, 3) a solution that guarantees that the resulting controller is both deadlock-free and maximally permissive, and 4) An ability to dynamically reevaluate the controller logic as the active part mix in the manufacturing system changes. The direct application of the R&W framework can involve a large search space as problem size grows. Extensions of our approach have addressed the problems of scalability, state space search, and execution time. This has been accomplished through the introduction of more effective state space search algorithms. These extensions and the relative efficiency of algorithms is also discussed and demonstrated in this chapter.