Incremental and Hierarchical Deadlock-Free Control of Discrete Event Systems with Variables: A Symbolic and Inductive Approach

Doktorsavhandling, 2015

Today's industry trend towards agile product development cycles and the ambition to shorten the time-to-market, represents an extremely competitive marketplace. This has driven the industry to use very complex and highly flexible manufacturing systems. In these systems, a large number of manufacturing operations need to be coordinated in order to fulfill the manufacturing requirements and assuring safety and deadlock-free behavior of the entire system. Supervisory control theory (SCT) is one of the formal methodologies that promises a systematic and automatic computation of controllers for coordination of manufacturing operations, more broadly, discrete event systems (DES). However, by increasing the number of operations, synthesizing controllers soon becomes unmanageable and can lead to state space explosion. This is one of the main reasons that the industrial acceptance of the SCT framework is scarce. In this thesis, we investigate this fundamental issue in the SCT framework and propose different approaches to cope with this problem. One of the challenges in synthesizing a controller for a DES is that we need to explore and examine all its possible behavior. However, for a DES with infinite behavior this process is not feasible due to time and memory limits. In this thesis, we propose a novel symbolic synthesis technique based on the IC3 algorithm, one of the most effective SMT-based model checking algorithm, for safe and maximally permissive supervisory control of infinite-state DES with variables. An evaluation of the proposed IC3-based technique on standard SCT benchmarks shows a radical improvement in computation of controllers for systems with large or infinite state space compared to BDD-based and SAT-based approaches. Furthermore, in practice, most of the manufacturing systems are distributed and often composed of several components. In this thesis, we also propose an incremental and hierarchical control architecture to obtain an effective and computationally efficient synthesis process for design of controllers for distributed DES. To this end, we exploit the efficiencies of symbolic techniques and synthesize controllers incrementally rather than at once for the entire system. Also, we use effective model abstraction techniques to abstract away unnecessary information to the synthesis process which, in turn, helps us to avoid building the entire state space of the systems. The computational effectiveness and practical usage of the introduced control architecture is illustrated by controller synthesis for a safe and deadlock-free coordination of operations in an industrial manufacturing cell.