Adaptive Output Regulation for Linear Systems with Unknown Sinusoidal Exogenous Inputs

Abstract

This dissertation investigates the output regulation problem (which is equivalent to the problem of asymptotic tracking and disturbance rejection when the reference inputs and the disturbances are generated by an autonomous differential equation, the so-called exosystem) for linear systems driven by unknown sinusoidal exosystems. Unlike previous researches, our ultimate goal is to achieve asymptotic regulation of the plant output to the origin for the sinusoidal exogenous signals (representing the reference inputs and disturbances) generated by the exosystems whose magnitudes, phases, bias, frequencies, and even the number of frequencies are all unknown. Here, the plant is linear time-invariant (LTI) single-input-single-output (SISO) systems (including non-minimum phase systems) without uncertainty. Before achieving the final control goal, we first start by considering an output regulation problem under the assumption that the number of frequencies contained in the exogenous inputs is known but magnitudes, phases, bias, and frequencies are unknown. To solve this problem, an add-on type output regulator with an adaptive observer is presented. The adaptive observer, based on the persistently exciting (PE) property, is used to estimate the frequencies of sinusoidal exogenous inputs as well as the states of plant and exosystem. Also, by add-on controller we mean an additional controller which runs harmonically with a preinstalled controller that has been in operation for the plant. When the desired performance of the preinstalled controller is not satisfactory, the add-on controller can be used. Some advantages of the proposed add-on controller include that it can be designed without much information about the preinstalled controller and it can be plugged in the feedback loop any time in operation without causing unnecessary transient response. Both simulation and experimental results of the track-following control for commercial optical disc drive (ODD) systems confirm the effectiveness of the proposed method. As the next step, we deal with the case where, as well as magnitudes, phases, bias, and frequencies, the number of frequencies contained in the exogenous inputs is unknown. To this end, a closed-form solution is given under the assumptions that the plant has hyperbolic zero dynamics (i.e., there is no zero on the imaginary axis of the complex plane), and that the number of unknown frequencies has known upper bound. In particular, the PE property is not necessary for the estimation of the unknown frequencies. For this, an adaptive observer is proposed to estimate the frequencies and the number of frequencies, simultaneously. This is important contribution, because, sufficient persistency of excitation is usually required since the unknown parameters are estimated by the adaptive control. Moreover, we propose a suitable dead-zone function with a computable dead-band only using the plant parameters to avoid the singularity problem in the transient-state and, at the same time, to achieve output regulation in the steady-state.