NARMAX Model-Based State-Space Self-Tuning Control for Nonlinear Stochastic Hybrid Systems

• Main Authors: Chu-TongWang, 王舉東
• Other Authors: Sheng-Hong Tsai
• Format: Others
• Language: en_US
• Published: 2010
• Online Access: http://ndltd.ncl.edu.tw/handle/00478786595080956982

博士 === 國立成功大學 === 電機工程學系碩博士班 === 98 === In this dissertation, a state-space self-tuning control scheme based on the nonlinear autoregressive moving average with exogenous inputs (NARMAX) model has been proposed and used to design the optimal fault-tolerant trajectory tracker for nonlinear stochastic hybrid systems with unknown system parameters, plant noises, measurement noises, and inaccessible system states. The contents of this dissertation include two main parts: i) The novel state-space self-tuning control scheme based on the rational NARMAX model has been proposed for the nonlinear stochastic hybrid systems mentioned above. For parameter estimation, the nonlinear parameter model needs to be transformed into a linear-in-the-parameter model via algebra operations. In the design process of state-space self-tuning control, the procedure to constructing the optimal state-space model in canonical observer form and the optimal trajectory tracker design have been developed for state estimation and self-tuner design. ii) For the nonlinear stochastic hybrid systems mentioned previously, the efficient state-space self-tuning control mechanism involving an active fault tolerance based on the polynomial NARMAX model has been addressed and utilized to achieve an active fault-tolerant pulse-width-modulated (PWM) trajectory tracker. A formula to computing the initial parameter used in recursive estimation algorithm has been derived via the observer/Kalman filter identification (OKID) algorithm and the optimal linearization method so as to not only reduce the identification process time, but also enhance the controlled system performances. A criterion is given to determine whether the system faults occur or not. For system faults, the method to initializing estimation error covariance matrices used in the modified Kalman filter algorithm has been also developed in order to prevent the performance degeneration due to the system faults.