Dynamic Simulation and prediction of drivers' attention distribution

Abstract

The distribution of driver’s attention is a crucial aspect for safe driving. The SEEV model by Wickens is a state of the art model that provides an easy but abstract way to estimate the distribution of attention for specific situations. The present paper presents an extension of the SEEV model, the Adaptive Information Expectancy (AIE) model. The AIE model is a sophisticated model of attention control, able to provide estimates based on a far more detailed simulation of human allocation of attention within a cognitive architecture. The AIE model relates attention directly to a task model, which is executed within the architecture. It is able to automatically measure task-dependent event frequencies and adapt its distribution of attention according to these frequencies. The AIE model was used to create a dynamic cognitive driver model. A driving simulator study with 21 participants has been conducted to evaluate the predictions of the driver model. Event rates for the primary driving task and an artificial secondary task have been varied, as well as the priorization of tasks. Both the SEEV and the AIE model provided estimates for percentage dwell times with similar quality, while the AIE model was able to provide estimates for further measure like gaze frequencies and link values.