A Study on the Performance Improvement of the Auto-Tuning PID Controller Using Gradient Method

경사도 기법을 사용한 PID 제어기의 성능 개선에 관한 연구

In this paper, we proposed a simple neural network-based parameter tuning algorithm, which could find the gradients of a certain performance index in the PID parameter spaces. In this process, we had to know the dynamics between input and output of the plant, and we used the Back Propagation Neural network to identify them. To make the parameter updating fast and smooth, we constructed the performance index as the sum of past N-squared plant errors, and applied a batch mode algorithm to update parameters. We performed several experiments with a DC Motor to show the validity of the proposed algorithm.