Neural network controller for flexible plate considering spillover effect on learning process

Abstract

We propose noncontact sensing and control methods of flexible plate materials handled by a robotic manipulator. A displacement sensor and a force/torque sensor are introduced to measure vibration. Based on the proposed neural network controller, which employs the feedback error learning method as the basic control architecture, this paper deals with the spillover effect to asymptotic convergence of learning. The conditions for asymptotic convergence of feedback error learning method for each trials are obtained. The influence of the vibration modes unmodeled on the conditions for asymptotic convergence is discussed. Based on the results obtained here, we present a new learning method to improve the control performance. The control system consists of a low pass filter and two neural networks. The learning method is to change the learning rate according to the convergence conditions. From simulation results, we show that the tracking performance is improved by using proposed learning method.