Adaptive Steering System for Improved User Experience

The steering system of an automobile serves as the initial point of contact for the driver and is a crucial determinant in the purchasing choice of the vehicle. The present steering system is equipped with a singular Electric Power Assisted Steering (EPAS) map, resulting in a consistent steering sensation during maneuvers conducted at both low and high velocities. Certain vehicles are equipped with a steering system that includes fixed driving modes that require manual intervention. This paper presents a proposed Machine Learning based Adaptive Steering System that aims to address the requirements and limitations of fixed mode steering systems. The system is designed to automatically transition between comfort and sports modes, providing users with the desired soft or hard steering feel. The system utilizes vehicle response to driver input in order to identify driving patterns, subsequently adjusting steering assist and torque automatically. The system consists of driving pattern recognition module which classifies driver intention into sport or comfort driving. The system functionality is supported by the existing vehicle CAN data executing in real-time. Decision tree algorithm is employed to classify the driving patterns, with a physical data based accuracy rate of 98%. Additionally, mode holding logic is introduced in the system to avoid fluctuations between the two modes. The system has been effectively validated through the execution of maneuvers such as double lane changes and high-speed cornering both in digital and physical domains.