USING NEURAL NETWORKS FOR ROBOT POSITIONING CONTROL

Automation of manufacturing processes often involves integrating some type of robot with the manufacturing system. Dynamic interactions between the robot and its environment introduce the importance of robot process capability (RPC) and the complication of robot control. In this research, a learning algorithm with a simple neural network was proposed not only for robot learning but also for simplifying control and improving the RPC in on-line processing. The proposed neural network has a modified two-layer counterpropagation network (MTL-CPN). Before the MTL-CPN learning algorithm was built, five algorithms were studied: Kohonen's neighborhood, DeSieno's conscience, error-range, and three-dimensional counterpropagation network (3D-CPN) learning algorithms. The simulation results show the effectiveness of the proposed algorithm for learning control.