本論文提出一適應性小波小腦模型(Wavelet-based-CMAC, WCAMC)網路,並將該網路應用到非線性系統鑑別、控制及影像壓縮上。WCMAC網路將小波函數引入小腦模型網路中,取代常用的高斯函數,此外也將T-S型式模糊推論的概念代入網路中,以提升網路對函數的近似能力,此改變使WCMAC網路在學習上不僅收斂快速,且使用之網路節點及參數個數較少。在非線性控制方面,以WCMAC網路為基礎,提出監督式小波小腦模型控制器(Supervisory Wavelet-based-CMAC Controller, SWC),SWC除了擁有WCMAC收斂快速的優點外,亦提出輔助控制器參數調整法則,讓控制器可以根據系統的變化即時調整,得到較佳的收斂速率與控制性能。此外針對非線性TORA(Translational Oscillations with a Rotational Actuator)系統,基於遞迴步階(Backstepping)之概念設計控制器,並結合SWC提出了監督式遞迴步階小波小腦模型控制器(Supervisory Backstepping Wavelet-based-CMAC Controller, SBWC),SBWC結合了遞迴步階控制器與SWC優點,即設計簡單、收斂速度快、擁有好的強健性。此外為顯現WCMAC之優點,本文將其應用於非失真影像壓縮,基於差值脈波編碼調變,以WCMAC網路建構一個智慧型預測器,來提升壓縮效率。最後經由模擬結果可看出,本文提出的方法在各方面確實有不錯的效果。
This thesis proposes an adaptive wavelet-based cerebellar model arithmetic controller (WCMAC) network, and its applications in nonlinear systems identification, control, and image compression. The Gaussian functions of traditional CMAC are replaced by wavelet functions. In addition, properties and advantages of a fuzzy TSK model are used to modify the activation functions of WCMAC. These modifications result small network structure and parameters, highly approximation capability, and fast convergence. For the nonlinear control problem, we propose a supervisory wavelet-based-CMAC controller (SWC) by using WCMAC network. The control performance is improved by the adaptability of SWC even the system is perturbed. Furthermore, by the concept of Lyapunov and backstepping, a controller design procedure is proposed for the nonlinear translational oscillations with a rotational actuator (TORA) system. Combine the advantages of backstepping controller and SWC a supervisory backstepping wavelet-based-CMAC controller (SBWC) is proposed. SBWC has the advantages such as simple design produce, fast convergence, and robustness. In order to show the advantage of WCMAC, it is used to develop an intelligent predictor for lossless image compression. Finally, several simulations are shown to demonstrate the effectiveness of the adaptive WCMAC system.