Adaptive control for applications with input saturation constraints or disturbances at poorly known frequencies

The research presented in this dissertation modifies an existing model reference adaptive control and direct adaptive disturbance rejection algorithm to better-accommodate real-life plants and disturbances. First, the theoretical result is extended to include input amplitude and rate saturation constraints while maintaining bounded output error and adaptive gains. The negative impact of enforcing control limitations is alleviated through an appropriate modification of the reference system command. Second, a heuristic approach is taken to improve the disturbance rejection effectiveness when the disturbance frequency is not well-known. Given the disturbance frequency (or waveform), the prior approach is able to mitigate persistent disturbances by adaptively compensating for unknown disturbance amplitude and phase. The framework is modified in this research to accommodate for disturbances occurring at coarsely-known frequencies, while continuing to compensate for its unknown amplitude and phase. This is done by including a phase locked loop to improve the frequency estimate on-line. As the frequency approximation improves, the controller is able to more effectively anticipate and reject persistent disturbances acting on the plant. The inclusion of both saturation constraints and on-line frequency estimation are illustrated using a detailed model and simulation of a powered ocean platform. The effectiveness of the adaptive output feedback control and disturbance rejection algorithm with realistic input amplitude and rate limitations is shown through its application to the spar platform for several challenging environmental and load disturbance scenarios. Furthermore, vortex shedding disturbances for near-constant relative water velocities across the spar occur at frequencies that cannot be precisely calculated. Inclusion of the phase locked loop is shown to improve the performance of the adaptive disturbance rejection algorithm through the use of both one and six degree-of-freedom simulations.