Event-triggering architectures for adaptive control of uncertain dynamical systems

In this dissertation, new approaches are presented for the design and implementation of networked adaptive control systems to reduce the wireless network utilization while guaranteeing system stability in the presence of system uncertainties. Specifically, the design and analysis of state feedback adaptive control systems over wireless networks using event-triggering control theory is first presented. The state feedback adaptive control results are then generalized to the output feedback case for dynamical systems with unmeasurable state vectors. This event-triggering approach is then adopted for large-scale uncertain dynamical systems. In particular, decentralized and distributed adaptive control methodologies are proposed with reduced wireless network utilization with stability guarantees. In addition, for systems in the absence of uncertainties, a new observer-free output feedback cooperative control architecture is developed. Specifically, the proposed architecture is predicated on a nonminimal state-space realization that generates an expanded set of states only using the filtered input and filtered output and their derivatives for each vehicle, without the need for designing an observer for each vehicle. Building on the results of this new observer-free output feedback cooperative control architecture, an event-triggering methodology is next proposed for the output feedback cooperative control to schedule the exchanged output measurements information between the agents in order to reduce wireless network utilization. Finally, the output feedback cooperative control architecture is generalized to adaptive control for handling exogenous disturbances in the follower vehicles. For each methodology, the closed-loop system stability properties are rigorously analyzed, the effect of the user-defined event-triggering thresholds and the controller design parameters on the overall system performance are characterized, and Zeno behavior is shown not to occur with the proposed algorithms --Abstract, page iv