抄録
This paper addresses the question of how to perform on-line training of multilayer neural controllers in an efficient way. At first, a plant emulator and a feedforward controller based on multilayer neural networks are described. Only a little qualitative knowledge about the plant is required. The controller must learn the inverse dynamics of the plant from randomly chosen initial weights. Basic control configurations are briefly presented. New on-line training methods, based on efficient use of memory-stored data and distinction between sampling and learning frequencies are proposed. One method, called direct inverse control error approach, is effective for small adjustments of the neural controller when it is already reasonably trained; another one, dubbed predicted output error approach, directly minimizes the control error and greatly improves convergence of the controller. Simulation results show the effectiveness of the proposed neuromorphic control structures and training methods.
