Design of Electric Power Steering System Identification and Control for Autonomous Vehicles Based on Artificial Neural Network

Abstract

Electric power steering (EPS) poses significant control challenges in autonomous vehicles due to their inherent complexity and non-linearity. This study explores the application of artificial neural network (ANN) to address these limitations. Two approaches are proposed: 1) an ANN-based identifier utilizing the backpropagation (BP) algorithm to learn the system’s non-linear dynamics, and 2) an ANN-based controller leveraging the Levenberg-Marquardt (LM) algorithm to improve control performance. Our findings demonstrate the efficacy of the proposed ANN-based BP algorithm in EPS system identification achieving over 99.6% accuracy in predicting EPS system dynamics compared to the traditional approach. Additionally, the LM-learned ANN-based controller aiming a faster response and precise reference tracking compared to the traditional controller method. These advancements underscore the potential of employing ANN methodologies to optimize EPS performance in autonomous vehicles.

Keywords

• Artificial neural network
• autonomous vehicles
• backpropagation
• electric power steering
• Levenberg-Marquardt
• proportional integral derivative controller