PhD Oral Exam - Nazanin Hashemi Attar, Electrical and Computer Engineering

When studying for a doctoral degree (PhD), candidates submit a thesis that provides a critical review of the current state of knowledge of the thesis subject as well as the student’s own contributions to the subject. The distinguishing criterion of doctoral graduate research is a significant and original contribution to knowledge.

Once accepted, the candidate presents the thesis orally. This oral exam is open to the public.

Abstract

Model uncertainty due to unknown dynamics or changes (such as faults) must be addressed in supervisory control design. Robust supervisory control, one of the approaches to handle model uncertainty, provides a solution (i.e., supervisor) that simultaneously satisfies the design objectives of all possible known plant models. Complexity has always been a challenging issue in the supervisory control of discrete-event systems, and different methods have been proposed to mitigate it. The proposed methods aim to handle complexity either through a structured solution (e.g., decentralized supervision) or by taking advantage of computationally efficient structured models for plants (e.g., hierarchical models). One of the proposed hierarchical plant model formalisms is State-Tree-Structure (STS), which has been successfully used in supervisor design for systems containing up to 10^20 states.

In this thesis, a robust supervisory control framework is developed for systems modeled by STS. First, a robust nonblocking supervisory control problem is formulated in which the plant model belongs to a finite set of automata models and design specifications are expressed in terms of state sets. A state-based approach to supervisor design is more convenient for implementation using symbolic calculation tools such as Binary Decision Diagrams (BDDs). In order to ensure that the set of solutions for robust control problem can be obtained from State Feedback Control (SFBC) laws and hence suitable for symbolic calculations, it is assumed, without loss of generality, that the plant models satisfy a mutual refinement assumption. In this thesis, a set of necessary and sufficient conditions is derived for the solvability of the robust control problem, and a procedure for finding the maximally permissive solution is obtained.

Next, the robust state-based supervisory framework is extended to systems modeled by STS. A sufficient condition is provided under which the mutual refinement property can be verified without converting the hierarchical model of STS to a flat automaton model. As an illustrative example, the developed approach was successfully used to design a robust supervisor for a Flexible Manufacturing System (FMS) with a state set of order 10^8.