Improving the Performance of Kalman Filter in PID Control Systems Using Neural Networks and Genetic Algorithms

Abstract

This article provides a study to enhance the performance of a control system, specifically the system relying on a PID (Proportional-Integral-Derivative) controller with a Kalman filter, by incorporating artificial intelligence techniques (genetic algorithm and artificial neural network) into the previous system. The research aims to analyze and evaluate the performance of the proposed system compared to the traditional PID controller with a Kalman filter. To achieve this, the system was applied to an Autonomous Underwater Vehicle (AUV) and simulations were conducted under ideal conditions (without noise) and in a marine operating environment (with noise).

Simulations were performed and the results were presented using MATLAB software. Simulation results of steering angle control for the vehicle (without noise) showed a decrease in Integral Absolute Error (IAE), overshoot, and settling time by 61%, 28%, and 78.4%, respectively. Meanwhile, simulation results (with noise) exhibited a decrease in IAE and overshoot by 94.9% and 23.5%, respectively. However, the settling time was not achieved in the PID model with a Kalman filter because the noise caused oscillations in the response function beyond the acceptable error range for settling time.

On the other hand, the settling time for the proposed algorithm with similar noise was comparable to the settling time without noise.