中文摘要 近年來,類神經網路已普遍運用在各領域,而運用在機械手臂的智慧型控制 理論,正受到人們廣泛的注意。由於機械手臂的高度非線性,使得難以依據其複雜的數學模式,來設計動態系統,但類神經網路的特性,是不需建立完整的數學模式,是易於掌握的。且其具有智慧型的學習及適應性,此一特性正適合來控制機械手臂。有鑑於此,在本篇論文中,我們利用倒傳遞類神經網路,來模擬機械手臂在平面繪圖。而在網路學習過程中,學習速率的選擇,將對模式之建立影響甚大,因此,必須建立一個有效的訓練模型,以利於類神經網路之學習。經實驗證實,學習速率的大小,確實對類神經網路學習系統的收斂影響甚大,而所設計的網路模式,具有學習及反應機械手臂系統特性之能力。所以,將類神經網路應用於機械手臂繪圖模擬,確實是可行的。
Abstract In recent years, Artifical Neural Network(ANN) have been applied in different field generally ,and the Intelligent Control theory which is used on Robot Arm are drawing people’s attention worldwidely .Due to the high nonlinear of the Robot Arm ,it’s hard to design the dynamic system according to its complicated mathematic model ,However the characteristic of ANN is not necessary to build the complete mathematic model. It’s easy to be controlled and it also has intelligent learning ability and adaptability.The characteristic is suit for controlling Robot Arm . According to this ,the writer utilize Back-Propagation Neural Network(BPNN) to simulate Robot Arm drawing on the plane .The selection of learning rate influences the build-up of model a lot , hence it is necessary to build up an effective training model to enhance the learning ability of ANN. Throughing empirical experiment , it is identified that the speed of leaening rate actually influences ANN learning system and the network model which is designed has the ability of learning and response to the characteristic of Robot Arm, therefore it is possible to implete the simulation of ANN applying to Robot Arm drawing.