Trajectory tracking using neural networks

Presents a method for tracking prespecified trajectories in industrial drive systems with a multilayered feedforward neural network. The method utilizes the backpropagation technique to learn feedback-error ranges, in which appropriate control actions can be generated from a lookup table. It can follow arbitrarily prescribed trajectories even when they are not present in the training phase. This approach is simple and practical for real-time implementation. Examples are included to demonstrate the effectiveness. The analogy between this scheme and a fuzzy logic control strategy is also investigated