From Security Enforcement to Supervisory Control in Discrete Event Systems: Qualitative and Quantitative Analyses

Abstract

Cyber-physical systems are technological systems that involve physical components that are monitored and controlled by multiple computational units that exchange information through a communication network. Examples of cyber-physical systems arise in transportation, power, smart manufacturing, and other classes of systems that have a large degree of automation. Analysis and control of cyber-physical systems is an active area of research. The increasing demands for safety, security and performance improvement of cyber-physical systems put stringent constraints on their design and necessitate the use of formal model-based methods to synthesize control strategies that provably enforce required properties. This dissertation focuses on the higher level control logic in cyber-physical systems using the framework of discrete event systems. It tackles two classes of problems for discrete event systems. The first class of problems is related to system security. This problem is formulated in terms of the information flow property of opacity. In this part of the dissertation, an interface-based approach called insertion/edit function is developed to enforce opacity under the potential inference of malicious intruders that may or may not know the implementation of the insertion/edit function. The focus is the synthesis of insertion/edit functions that solve the opacity enforcement problem in the framework of qualitative and quantitative games on finite graphs. The second problem treated in the dissertation is that of performance optimization in the context of supervisory control under partial observation. This problem is transformed to a two-player quantitative game and an information structure where the game is played is constructed. A novel approach to synthesize supervisors by solving the game is developed.

The main contributions of this dissertation are grouped into the following five categories. (i) The transformation of the formulated opacity enforcement and supervisory control problems to games on finite graphs provides a systematic way of performing worst case analysis in design of discrete event systems. (ii) These games have state spaces that are as compact as possible using the notion of information states in each corresponding problem. (iii) A formal model-based approach is employed in the entire dissertation, which results in provably correct solutions. (iv) The approaches developed in this dissertation reveal the interconnection between control theory and formal methods. (v) The results in this dissertation are applicable to many types of cyber-physical systems with security-critical and performance-aware requirements.