DEADLOCK AVOIDANCE IN FLEXIBLE FLOW SHOPS WITH LOOPS Paul Valckenaers a , Hendrik Van Brussel a , Sven Brueckner b , Jo Wyns a and Patrick Peeters a a Katholieke Universiteit Leuven, Mechanical Engineering Department, Leuven, Belgium b Center for Electronic Commerce, ERIM, Ann Arbor, USA Abstract: This paper presents a deadlock avoidance scheme that was developed during the Mascada Project—Esprit LTR Project 22728. The deadlock problem occurs in a car body painting shop. From this application, an abstract problem was derived for which a deadlock avoidance method was developed. The paper presents the deadlock avoidance scheme and its correctness proof. In addition, implementation issues are introduced and solutions are discussed. Keywords: Agents, Deadlock, Manufacturing Systems, Intelligent Manufacturing Sys- tems, Flexible Manufacturing Systems. 1. INTRODUCTION This paper presents the deadlock avoidance scheme that was developed during the Mascada Project (Es- prit LTR Project 22728) sponsored by the European Commission. The paper first introduces the applica- tion in which the deadlock problem occurs. Next, it describes the abstract problem derived from this ap- plication for which a deadlock avoidance method was developed. Third, the manuscript introduces the deadlock avoidance scheme and its correctness proof. Fourth, implementation issues are introduced and solutions are discussed. Finally, the paper lists future research aspects in this context. 2. THE APPLICATION The deadlock avoidance mechanism, presented in this paper, targets a specific production system—an existing car body paint shop; the presented deadlock handling method nevertheless is suitable for an entire class of systems. The manufacturing control system, which is enforcing the deadlock avoidance mecha- nism, is responsible for all routing decisions of car bodies moving through a paint shop. This system is composed of the following types of manufacturing resources: • Longitudinal unidirectional conveyors • Lateral unidirectional and bi-directional convey- ors • Turning tables • Lifts • Sorting buffers (automatic storage and retrieval subsystems) • Processing units (car body painting) These manufacturing resources are arranged in a complex topology, which has emerged from numer- ous upgrades that occurred during the painting shop's lifetime. The paint shop comprises more than 400 of these resources, paints more than 1000 cars per day and fills a building comprising six floors. In all, it presents a formidable control problem, especially