A New Approach for Steering Wheel Angle Prediction

Approaches to monitor drivers' attention generally involve evaluating the difference between the predicted degree of the steering wheel angle and the actual one. The effectiveness of this solution mainly depends on the quality of the prediction model of the steering wheel angle. This paper introduces a new approach to selecting the information which characterizes the dynamic state of the vehicle in order to integrate it into the prediction model. The proposed method relies on the analysis of a set of specific driver reactions which are assumed to characterize the steering behavior and which mainly depend on the dynamic states of the vehicle (DSV). The fundamental issue of this approach is to propose a generic structure of a prediction model taking into account this information. This paper discusses the first results obtained by this method when applied to experiments using a driving simulator.