Identification of a Linear Parameter Varying Driver Model for the Detection of Distraction

Abstract

This study investigates a real-time identification of a LPV Cybernetic Driver Model (CDM), using the Unscented Kalman Filter (UKF). Unlike an iterative identification, dealing with packets of input-output data, the method considered is recursive identification dealing with the actual input-output data, making explicit the parametric changes along time. The two identification schemes are implemented and the cybernetic driver model capability to predict or to estimate the driver steering was evaluated. The possibility of the recursive identification, when applied to the cybernetic model, to help for distraction diagnosis is finally evoked. The variations of the driver model parameters are shown to be significantly associated with driver distraction.