Multiple Model Adaptive Control Of Parameter Uncertain Systems

In recent decades, the control problem of systems with uncertainty gains more and more attention. Adaptive control scheme, which is one of the effective ways to deal with uncertainty, is commonly used and a lot of related work has been done. As the performance of the system becomes more and more important, many ideas have been proposed to improve the classical adaptive control scheme. Among them, the idea of using multiple models gains great interests due to the capability of dealing with large uncertainty. In this dissertation, we focus on multiple model adaptive control scheme with switching and second level adaptation in dealing with nonlinear and quantized-linear parameter-uncertain systems. Both the design methodology and performance improvement mechanism are discussed.As for the multiple model adaptive control scheme with switching, the adaptive control of nonlinear system with output feedback is under consideration. The classical controller and parameter update law are designed based on Luenberger observer. As the controller with state estimation is not precise, a compensator is proposed to provide extra information, which is capable of reducing the side-effect of state estimation error as much as possible. Further, the multiple adaptive observer scheme is proposed to improve the performance of state estimation and system output. The dwell-time is designed according to the energy of the identification models to assure the stability of the whole system.When the state feedback is quantized, the adaptive control problem of linear uncertain system is not ever considered. As the quantized state feedback is not continuous, the classical adaptive control scheme fails in some sense. In order to avoid the complexity of controller and parameter update design with quantized state feedback, the multiple model adaptive control scheme with switching using fixed identification models is proposed. The stability of the error system of exact-matching model is analyzed and the quantizer is designed. Also, the hysteresis switching scheme is used to assure the stability of the system.As for the multiple model adaptive control scheme with second level adaptation, the fundament principle and design methodology are discussed in detail. Also, the performance improvement over classical adaptive control scheme is demonstated in both theoretical analysis and experiments. Meanwhile, the multiple model adaptive control scheme with second level adaptation using fixed identification models is considered, and the stability is obtained based on the convex combination of identification errors.Up to now, the research about multiple model adaptive control with second level adaptation scheme mainly focus on linear system, and it is extended to nonlinear systems for the first time in this dissertation. A special form of identification model is designed according to the structure of strict-feedback nonlinear system, and the overall system is considered to obtain the controller. Also, the second level adaptation is extended to nonlinear systems based on the character of exponential convergence.Further, when the identification models are adaptive, the controller and parameter update laws are obtained based on the analysis of all the identification models and the system, which assures the stability of the system. While when the identification models are fixed, the controller design is much simpler and it can deal with time-varying systems. Also, the stability is maintained according to the analysis of convex combination of identification errors.Finally, some simulations are carried out to demonstrate the feasibility and efficiency of the proposed methods.