Supervisory control of fuzzy discrete event systems with applications to mobile robotics

Abstract

Fuzzy Discrete Event Systems (FDES) were proposed in the literature for modeling and control of a class of event driven and asynchronous dynamical systems that are affected by deterministic uncertainties and vagueness on their representations. In contrast to classical crisp Discrete Event Systems (DES), which have been explored to a sufficient extent in the past, an in-depth study of FDES is yet to be performed, and their feasible real-time application areas need to be further identified. This research work intends to address the supervisory control problem of FDES broadly, while formulating new knowledge in the area. Moreover, it examines the possible applications of these developments in the behavior-based mobile robotics domain. An FDES-based supervisory control framework to facilitate the behavior-based control of a mobile robot is developed at first. The proposed approach is modular in nature and supports behavior integration without making state explosion. Then, this architecture is implemented in simulation as well as in real-time on a mobile robot moving in unstructured environments, and the feasibility of the approach is validated. A general decentralized supervisory control theory of FDES is then established for better information association and ambiguity management in large-scale and distributed systems, while providing less complexity of control computation. Furthermore, using the proposed architecture, simulation and real-time experiments of a tightly-coupled multi-robot object manipulation task are performed. The results are compared with centralized FDES-based and decentralized DES-based approaches. -- A decentralized modular supervisory control theory of FDES is then established for complex systems having a number of modules that are concurrently operating and also containing multiple interactions. -- Finally, a hierarchical supervisory control theory of FDES is established to resolve the control complexity of a large-scale compound system by modularizing the system vertically and assigning multi-level supervisor hierarchies. As a proof-of-concept example to the established theory, a mobile robot navigation problem is discussed. This research work will contribute to the literature by developing novel knowledge and related theories in the areas of decentralized, modular and hierarchical supervisory control of FDES. It also investigates the applicability of these contributions in the mobile robotics arena.