模糊控制以及模糊類神經網路在推廣型卡爾曼濾波器之應用

Abstract

本文對於模糊控制以及模糊類神經系統，做了深入的研究與探討。 在模糊控制方面，我們首先推導模糊控制器與傳統PD (或PI)控制 器之等效性；接著，我們還提出了一個無穩態誤差響應的自調式 PD模糊控制器 另一方面，我們發展了兩個模糊類神經系統： NFNN 及 FNNS，用以簡化模糊類神經網路的複雜度。此外，我們還 發展出一個結合此二系統特點的合成方法，它能在不需事前的專 家知識情況下，有彈性地鑑別並簡化模糊類神經網路的架構。最 後，我們利用模糊類神經網路建立了一個離散推廣型卡爾曼濾波 器用以估測非線性系統的狀態。 In this thesis we do the work of research about the fuzzy control and fuzzy-neural systems. In fuzzy control, we first propose a fuzzy logic controller which is equivalent to the classical PD (or PI) controller. A PD-like self-tuning fuzzy controller is then presented that yields zero steady-state responses. On the other hand, two fuzzy-neural systems, the NFNN and FNNS, are developed for reducing the complexity of a fuzzy neural network. Also, a synthesis method combining the advantages of NFNN and FNNS is explored to flexibly identify a fuzzy-neural-network structure without prior expert knowledge. Finally, we construct a discrete extended Kalman filter by using fuzzy neural networks.