智慧型ANFIS 馬達控制器設計應用於運動物體之追蹤

Abstract

本論文目的在於利用智慧型ANFIS控制器以及利用Runge-Kutta 的方法去建立馬達的離散模型，並藉由倒傳遞學習演算法去調整找出ANFIS控制器中近似最佳的參數值，當取樣時間太大或者模型有快速的動態變化，那這樣一階的線性近似將不是一個適當方法去估測下一個狀態，為了解決這樣的問題，可以使用較高階的Runge-Kutta 的公式，但相對的也會導致較複雜的適應性網路且會使得學習過程變慢。而在此論文中，假如模型變的複雜，則使用二階Runge-Kutta的公式而言，是相對的較簡單，而對於四階Runge-Kutta的公式而言，又會使得適應性網路變的太複雜，所以在本論文中所使用的為三階Runge-Kutta的公式去建立馬達的模型，然後在做馬達的定位到零度的學習，最後在利用學習到的控制器參數去做單個馬達角度的追蹤以及控制三軸馬達的仿人眼系統的運動物體追蹤之應用。

The objective of the thesis is to use the intelligent ANFIS controller and to build the discrete PM DC motor by using the Runge-Kutta method, then use the back propagation learning algorithm to find the near-optimal parameters in ANFIS based controller. When the sampling time h is too large or the plant has fast dynamics, the linear approximation may not be a suitable estimate of the next state [5]. In order to solve this problem, the so-called Runge-Kutta method with higher order can be used to implement the plant block. However, the use of higher order Runge-Kutta method often results in more complicated adaptive network and slows down the learning process accordingly. In this thesis, there are used the third order Runge-Kutta formula because the second order is too easy if the model become complex and the fourth order will result complicated adaptive network for the modeling. Then by using the regulation learning of motor to find the parameters in controller applied to angle tracking of one motor and to track the moving object by controlling the humanoid vision system which is three axis actuated by three motors.