| dc.description.abstract | Multi-agent systems (MASs) have been widely recognized as a key way to model, analyze, and engineer numerous kinds of complex systems composed of distributed agents. The aim of this dissertation is to study control design for leader-follower MASs such that a group of followers can track a specified leader via distributed decision making based on distributed information. We identify and consider several critical problems that have stood in the way of distributed tracking control synthesis and analysis. Specifically, they include: 1) limited information access by the followers to the leader, 2) effects of external disturbances, 3) complicated dynamics of agents, and 4) energy efficiency. To overcome the first three problems, we take a lead with the design of distributed-observer-based control, with the insight that distributed observers can enable agents to recover unknown quantities in a collective manner for the purpose of control. To deal with the fourth problem, we propose the first study of MAS tracking control conscious of nonlinear battery dynamics to increase operation time and range. The dissertation will present the following research contributions. First, we propose the notion of designing distributed observers to make all the followers aware of the leader's state and driving input, regardless of the network communication topology, and perform tracking controller design based on the observers. Second, we further develop distributed disturbance observers and observer-based robust tracking control to handle the scenario when all the leader and followers are affected by unknown disturbances only bounded in rates of change. The third contribution lies in treating a leader-follower MAS with high-order, nonlinear dynamics. Assuming the availability of very limited measurement data, we substantively expand the idea of observer-based control to develop a catalog of distributed observers such that the followers can reconstruct large amounts of information necessary for effective tracking control. Finally, we propose a distributed predictive optimization method to integrate onboard battery management with tracking control for long-endurance operation of an electric-powered MAS. The proposed dissertation research offers new insights and a set of novel tools to enhance the control performance of leader-follower MASs. The results also have a promise to find potential applications in other types of MASs. | |
