Neural controllers for electrical power steering systems

Modern electric cars require electrical power steering systems (EPAS). Many control algorithms where employed in this field. Some of these controllers exhibit robustness and stability problems for certain road conditions. Neural networks are known for their ability to imitate systems and stay stable if operation conditions change. In this paper we use neural controllers to imitate the H∞ controller we have already designed to control the EAPS system. We collect the H∞ performance signals and use them as training data for the suggested neural controllers. Fuzzy Adaptive Resonance Theory (fuzzy ARTMAP) and back propagation neural controllers are used in this paper to do the control act. The performance of each controller is recorded for comparison purposes.