Autonomous Robots 9, 59–69, 2000 c  2000 Kluwer Academic Publishers. Manufactured in The Netherlands. Moore Automata for the Supervisory Control of Robotic Manufacturing Workcells A. RAMIREZ-SERRANO, S.C. ZHU AND B. BENHABIB Computer Integrated Manufacturing Laboratory, Department of Mechanical and Industrial Engineering, University of Toronto, 5 King’s College Road, Toronto, Ontario, Canada, M5S 3G8 beno@mie.utoronto.ca Abstract. The potential of flexible-manufacturing workcells (FMCs) to produce a family of parts in many possible orders of operations and choices of different machines is advantageous. Despite intensive research on the theoretical control of discrete-event systems (DESs), however, current techniques can still only be used for the supervisory control of simple cells. In this paper, a novel modeling and control synthesis technique is presented for FMCs that allow part-routing flexibility. Our proposed methodology combines Extended Moore Automata (EMA) and Controlled-Automata theories to synthesize supervisors for such FMCs. Keywords: flexible manufacturing workcells, supervisory control, Moore automata, alternative processing routes 1. Introduction A flexible-manufacturing workcell (FMC) is an inte- grated system that consists of an arrangement of different-purpose machines linked with automated handling devices under the control of a supervisor (Benhabib et al., 1989). In practice, such discrete- event-systems (DESs) exist in a variety of configura- tions and degrees of complexity. They are capable of fabricating products under changing production pat- terns by using part-routing flexibility while possibly sharing resources. There exists no common agreement as to which is the most effective DES modeling technique, particularly for the purpose of FMC control (Sobh and Benhabib, 1997). Petri nets (PNs) and Controlled-Automata the- ory, however, are the two formal methods proposed by a large number of researchers (e.g., D’Souza and Khator, 1994; Lauzon et al., 1997). Two general types of PN methods have been proposed: those of the first type impose restrictions on situations that can be mod- eled and do not include the use of shared resources and complex ordering and selection criteria on part routes; and, those of the second type use extensions that al- low greater flexibility with respect to situations that can be modeled, but the models obtained are large and normally cannot be formally analyzed. In contrast, the attractiveness of Controlled-Automata theory over PNs lies on its high expressive power and on the primary characteristic that the supervisory design technique guarantees the control strategy to be correct (deadlock- free) by construction (Lauzon et al., 1997; Ramadge and Wonham, 1987; Williams, 1996). The range of applicability of the PNs and Controlled- Automata can be limited to manufacturing systems that are restricted in size and that can process parts with fixed linear routes. A few studies that did address the routing problem proposed techniques which can only control the flow of parts based on (pre-selected) pro- cessing routes (e.g., Reveliotis, 1998; Lawley, 1998; Byrne and Chutima, 1995). When specific parts need to be rerouted, a new route from the parts’ route space is selected and the supervisor is re-synthesized. Further, although formal control methods of FMCs have received wide attention within academia, their im- plementations are scarce. In Lauzon et al. (1997) and Jafari and Boucher (1994), the manufacturing workcell was controlled using Programmable-Logic-Controllers (PLCs), which were programmed using Controlled- Automata theory. In Balemi (1992) and Zhou and DiCesare (1993), a host-computer was used as a super- visor for the control of the DES using Petri-Nets. In