On Design and Analysis of Cyber-Physical Systems with Strategic Agent

In contrast to traditional CPS where a designer can specify an action plan for each agent, in CPS with strategic agents, every agent acts selfishly and chooses his strategy privately so as to maximize his own objective. In this dissertation, we study problems arising in the design and analysis of CPSs with strategic agents.;We consider two classes of design problems. In the first class, the designer utilizes her control over decisions and resources in the system to incentivize the agents via monetary incentive mechanisms to reveal their private information that is crucial for the efficient operation of the system. In particular, we consider market mechanism design for the integration of renewable energy and flexible loads into power grids. We consider a model that captures the dynamic and intermittent nature of these resources, and demonstrate the advantage of dynamic market mechanism over static market mechanisms that underly the existing architecture of the electricity markets. In the second class of design problems, the designer utilizes her informational advantage over the agents and employ informational incentive mechanisms to disclose information selectively to the agents so as to influence the agents' decisions. Specifically, we consider the design of public and private information disclosure mechanisms in a transportation system so as to improve the overall congestion.;We also study the analysis of CPS with strategic agents as a stochastic dynamic game of asymmetric information. We present a set of conditions sufficient to characterize an information state for each agent that effectively compresses his private and common information over time. This information state provides a sufficient statistic for decision-making purposes in strategic and non-strategic settings. Accordingly, we provide a sequential decomposition of the dynamic game over time, and formulate a dynamic program that enables us to determine a set of equilibria of the game. The proposed approach generalizes and unifies the existing results for dynamic teams with non-classical information structure and dynamic games with asymmetric information.