On the visual search of human drivers: an eye-tracker based study at a driving simulator

To improve the acceptance of ADAS systems and autonomous driving logics, it is fundamental to develop accurate models able to reproduce the behaviour and the perception of human drivers. To this end, a driving simulator study was performed with the aid of an eye-tracker device. The sight behaviour of 25 drivers was analysed in two typical urban scenarios. Namely, the scenarios are a straight road and a four-way crossroad. The sequence of two-way fixations was analysed, together with the dwell time associated to a given area of interest of the scenario. The research highlighted that, for the straight road, the most visited areas were the central road and the dashboard. The typical dwell time associated to them was the longest, as well. For the crossroad, a relevant number of fixation counts was associated also to the left and right road areas of interest. Finally, a Markov chain model aimed at reproducing the gazing pattern shown by the drivers was proposed.