A Framework for the Integration of Dynamic Observation, Control and Knowledge in Discrete-Event Systems

点击翻译按钮获取中文摘要

My research focuses on developing mathematical models to represent systems whose behaviour can be thought of as sequences of events or actions in order to automate problem-solving and test potential solutions in such discrete-event systems. The goal of the modeling is to generate controllers to meet given specifications. Controllers, which may be software, hardware or human strategies, must make decisions about which events to disable at various points in time and they must make their decisions based only on partial observations of the system. Prior work in the field uses models where controllability and observability of an event are static properties. In contrast, my proposed research aims to model dynamic control and dynamic observation.****Furthermore, just as humans make decisions based not only on their direct observations of the physical world, but also based on inferences they make about what other people know, so too should automated agents make decisions based on what they know about what other agents know. Consequently, my research will focus on the interaction between observation, control, and "knowledge". I will incorporate formal models of agent knowledge into my framework. ******Fundamental shifts in the way we view observation and control are anticipated in the proposed research program. By incorporating dynamic observation and control, a supervisory control theory model of service composition will be developed for service-oriented computing applications. My research will also have an impact in decentralized distribution systems, for example in electricity markets where the status quo will no longer be viable in the future, when autonomous agents (homes, electric vehicles, private generators) will produce and consume electricity. My research into knowledge-based models for control can be applied to strategies where individual agents are guided by incentives to act in a way that produces some overall desired results. ******Most ambitiously, my research will be applied to solve computer security problems in which adversarial agents have only partial observation of a system and yet try to exert control over the system to glean secret information. With more accurate models of discrete-event systems-based on dynamic observation and dynamic control and using knowledge-based reasoning-we can develop controllers that prevent adversaries from being able to discern information that is intended to be kept private. I will investigate scenarios where the attacker can control and change the system's behaviour. I will consider possible responses a controller can take. I hope to determine whether it is more advantageous to restrict the behaviour of the system or to prevent the attacker from controlling some events. This will have an impact, for example, on Canadian banking and credit card companies, where protecting the privacy of information is of paramount concern.***The research will train 10 HQP.**