Various types of driver models have been proposed in the literature, such as taxonomic, functional, and motivational. Recently, the promising Michon model was extended, leading to the widely used GADGET matrix. Nevertheless, the correlation of such models to actual road accidents and their causes has so far been weak. In addition, the use of those models for predicting driver behavioural adaptation has not proven feasible. This paper introduces a new concept for modelling driver's performance, that of DRIVABILITY. DRIVABILITY is defined as a combination of permanent and temporary factors that affect a driver's performance. Furthermore, this paper proposes a DRIVABILITY index and a methodology to measure it, in order to move from over-simplistic, hierarchical modelling to a multi-dimensional sphere. The usability of the newly proposed concept is shown through its application in three different example cases, including a system monitoring driver hypovigilance, a system for driver basic training, and an elderly driver's assessment scheme. (Author/publisher)
Abstract