以PC-based實現CMAC類神經網路控制系統及其兩軸控制之模擬

Abstract

本論文的主要目地是以個人電腦(PC)設計小腦模組關節控制器(CMAC)控制 系統﹐結合傳統PID控制器形成完整控制架構。所有的控制訊號將完全由 PC產生。近年來由於PC的工作速率大為提升﹐使得PC在各領域的運用更加 受到重視。對於類神經網路所須的大量計算時間也可大幅縮短﹐因此以 PC-based的控制設計是一個相當不錯的選擇。我們用修正的CMAC學習架構 以軟體模擬CMAC的控制結果﹐以了解CMAC在控制系統中伴演的角色﹐過去 CMAC發散的現象已經改善。接下來我們以PC經由數位/類比轉換電路送出 控制訊號､從類比/數位轉換電路抓回轉動位移､再以取樣時間電路固定 取樣時間完成整個控制A型臂第一軸的控制架構﹐最後並對實驗結果加以 討論。此外我們還模擬CMAC控制系統控制兩軸的情形﹐以對將來的多軸控 制做準備。我們設計一個六維輸入､二維輸出的CMAC學習架構以同時學習 兩個相關聯控制對象的系統反函數﹐獲得了不錯的成果。 The object of this paper is to design a CMAC control system by PC-based design combined by traditional PID controller. All control signal will be generated by PC. Due to the speed rising of PC recently﹐The application of PC is more important in several field. The large counting time of Neural Network can be reduced. Therefore, PC-based control system is a good choose.We use the correcting CMAC learning rule to simulate a CMAC control system, in order to realize what CMAC can play in this structure. The phenomenon of CMAC divergence have been improved. And, System generate a control signal to motors by D/A converter circuit from PC, get the locomotion by A/D converter circuit, and fix the sampling time by sampling clock circuit in order to control A-type machine arm. Finally, we show the result and diccussions.We still simulate the CMAC control system used in two coupling articulation .We generate a 6-input､2-output structure to learn the system inverse function of two coupling plant and control them, and get a good performance.