Optimal Control of Nonlinear Systems Using Generalized Kernel Models

• Main Authors: Hao-han Hsu, 徐皓涵
• Other Authors: Jason J.H. Tsai
• Format: Others
• Language: en_US
• Published: 2007
• Online Access: http://ndltd.ncl.edu.tw/handle/57145416631898313805

碩士 === 國立成功大學 === 電機工程學系碩博士班 === 95 === An optimal controller for nonlinear system identified by generalized kernel model is proposed in this thesis. EP algorithm is used to tune the elements of the diagonal covariance matrix for the kernel regressors. By applying EP algorithm on the kernel model, the error between regressors and the actual nonlinear system could be minimized and the terms of the regrssors could be reduced as well. By applying the optimal linearization approach, a generalized kernel model with estimated states for nonlinear continuous–time stochastic systems can be constructed for the optimal control. The proposed method enables development of digitally implement advanced control algorithm for nonlinear stochastic hybrid systems.