Driver-Steering Interaction Model and Evaluation of Driver Interaction Strategies in Assisted Driving

Advanced Driver Assistance System (ADAS) aim to enhance the driving experience for the driver. However, they can not make their contribution if drivers do not accept them and are not willing to purchase or use them. If the interaction between driver and ADAS can be described and effectively analyzed, this greatly improves the understanding of the conflict and synergy of driver and automated system. This paper provides a novel model of driver-steering interaction. The model estimates driver's desired torque target as well as potential conflicts of interest between driver and active ADAS. We apply the Variable Coefficient (VC) method to identify the time-varying parameters of the model taking into account physical constraints. The model is validated on real driving data with 58 drivers on a racing track. Based on the driver interaction model, we apply supervised machine learning algorithms using the MATLAB Classification Learner App to distinguish five interaction strategies drivers applied when interacting with the assistance system.