![](/multimedia/files/WSEAS Logo.jpg)
Plenary Lecture
Comparison of Four Neural Network Based Controllers Methods of Robot Trajectory Control
Professor Şahin Yildirim
Mechatronics Engineering Dept.
Engineering Faculty
Erciyes University
TURKEY
E-mail: sahiny@erciyes.edu.tr
Abstract: This lecture describes four methods for robot trajectory control. These methods are a standard PID controller, a Computed Torque Method (CTM), a neural network based inverse controller and a neural network based Internal Model Controller (IMC). The IMC is investigated as an alternative to the basic inverse control scheme that is difficult to implement. The results presented show the superior ability of the proposed neural network based IMC scheme at adapting to changes in the dynamic parameters of the robot. The proposed recurrent hybrid neural network has two advantages: the recurrent connections provide an inherent dynamic memory which facilitates modelling and the hybrid hidden layer enables real systems comprising linear and non-linear parts to be modelled accurately. The results of applying the recurrent hybrid network to the simulated control of the trajectory of a planar robot arm have demonstrated its superiority over feedforward and purely linear or non-linear networks. The proposed neural control scheme has also been shown to perform better than the conventional computed-torque and PID schemes.
Brief Biography of the Speaker: Dr YILDIRIM has received the MSc degree in Mechanical Engineering in 1990 from Erciyes University, Turkey. He has received his PhD degree in 1998 from System Engineering Department, Cardiff University, UK. He was established Mechatronic Engineering Department in 2005, Kayseri, Turkey. He is head of the Mechatronic Engineering Department and head of Foreign Relations Office of Erciyes University. He is lecturing the control theory, robotics, neural network applications in engineering at Erciyes University, Turkey. He has been supervised some national and international projects such as mobile nurse robot design and control, active vehicle system design and control, 2 legged walking robot control. He has been reviewed more than 100 papers. He is also author of more than 140 journal and conference papers. He is currently supervising more than 30 PhD and Msc students. His research area is consisted of Neural Network, Control Theory and Applications, Mobile and Industrial Robots, Vehicle Dynamics and Control.