Adaptive Control Of System With Unmodeled Dynamics

In actual control system, there are always some unmodeled dynamics components. When dealing with these unmodeled dynamics components, the designed system cannot achieve expected effect under an ideal condition in which unmodeled dynamics components are simply ignored. Therefore, in order to obtain a better application result in practical system from controllers designed for the model, unmodeled dynamics components have to be taken into consideration when designing a controller. However, unmodeled dynamics components with a high degree of uncertainty make it difficult for controller design, which is the research topic of this thesis.This thesis begins with an analysis of the existence of unmodeled dynamics from the perspective of system identification. Based on the analysis, methods for handling unmodeled dynamics and ideas for further work, namely force modeling method and signal compensation method, are then proposed and briefly discussed. Then, two kinds of potentially existing unmodeled dynamics components (multiply and add) are processed, using model reference adaptive control. Firstly, when considering unmodeled dynamics with only multiply part under corresponding assumptions, model reference adaptive control scheme is designed by separately taking advantage of local parameter optimization and the second of Lyapunov stability theory. The designed controller is, to some extent, resistant to unmodeled dynamics with multiply part. Later, targeted system with the two unmodeled dynamics parts is researched, using adaptive control method,σ-correction hybrid adaptive control is also introduced to realize a rapid adjustment of controlling parameters.For a better handling of unmodeled dynamic component in practical controlling process and based on a non-parameter model (NARX model), one order linear part is extracted and processed with common Linear feedback control method. Considering the remained part in NARX model as unmodeled dynamic component, this thesis also designs a adaptive control scheme, which does not rely on systematically mathematic modeling and is used as outer loop compensation control scheme to resist unmodeled dynamic. Then, the outer loop compensation control scheme is proved to be stable and convergent and extended to process the entire NARX model, making it more universal and convenient to utilize.Finally, using MATLAB, this thesis conducts a simulation for an actual vehicle lateral semi-active suspension control system. For the fact that there are many unknown or inestimable unmodeled dynamics components in vehicle lateral motion system, model reference adaptive control is applied. Then outer loop compensation control scheme based the non-parameter is added to offset unmodeled dynamics components in the system and restrain influence of uncertain parameters or random disturbance on the system, verifying stability and robustness of the proposed controlling method in this thesis.