Observer-Based Digital Redesign Tracker for Nonlinear Sampled Data Systems via a Hybrid Design of EP-Based Iterative Learning Control and Linear Quadratic Analog Tracker

• Main Authors: Yan-Yi Du, 杜彥頤
• Other Authors: Sheng-Hong Tsai
• Format: Others
• Language: en_US
• Published: 2007
• Online Access: http://ndltd.ncl.edu.tw/handle/27812654937025861757

碩士 === 國立成功大學 === 電機工程學系碩博士班 === 95 === In this thesis, a novel observer-based digital redesign tracker, whose parameters are given by a hybrid design of EP-based iterative learning control and linear quadratic analog tracker, is proposed to resolve the output tracking problem for a class of nonlinear sampled data systems. The tracker possesses good tracking performance in transient state response as well as in steady that when an appropriate sampling time is given. Furthermore, its tracking result is better than that of an original digital redesign tracker at a large sampling time. Firstly, the ideal system state is obtained from the hybrid design of iterative learning control (ILC) and linear quadratic analog tracker (LQAT). Then, according to the optimal linearization algorithm, the nonlinear system will be converted into the linear model by the state as operating point at each sampling time. Finally, these models will be applied to design the digital redesign tracker which preserves the perfect tracking performance of the analog controller. In addition, an observer, whose design idea is derived from the method of the linear quadratic regulator (LQR), will be applied in the digital redesign tracker, when some system state could not be acquired. Furthermore, in order to improve the learning convergence of ILC, evolutionary programming will be applied to search for the optimal learning gain of ILC; three improved subjects of ILC, the initial control input decision, the observer design by optimal linearization algorithm and LQR, and the improved update learning rule for the second order derivative system, whose product matrix of C and B is not full rank, will be presented in this thesis. Several MIMO illustrative examples will be presented to demonstrate the effectiveness of the proposed methodology.