A perception module for car drivers visual strategies modeling and visual distraction effect simulation

International audience; The aim of this research is to develop and implement a computational model able to simulate drivers' visual strategies in dual-task conditions and to investigate visual distraction effects. The modeling approach supporting this research is based on a cognitive model of the car driver so-called COSMODRIVE, focused on mental representations simulation (i.e. situational awareness) and implemented on a virtual platform (so-called SiVIC). In this framework, a module of the visual perception is needed for model interaction with the virtual road environment. This perception module is indeed the ''entry point'' for road scene analysis and decision-making. This modeling work is based on empirical data collected among 20 human drivers. Experiments have been designed in order to study visual distraction impact of driver's visual scanning. During this experiment, participants drove a car on simulator and answered at the same time to a visual secondary task. Visual strategies were collected through video films, and have been analyzed in order to model visual distraction effect. Different visual strategies for alternative scanning of the road scene and the pictogram screen (i.e. distraction in the cockpit) have been identified among our set of drivers. These empirical results have been then used to develop a visual perception module liable to dynamically explore the road environment, according to visual distraction effects. Through these empirical data, the perception module has been developed in order to reproduce these visuals strategies of human when a visual distraction occurs. The module is primarily based on a enhancement of a pre-existing virtual camera model in SiVIC platform, now used as a model of the human eye. The movement of gaze and target tracking can be simulated with the rotation of the camera depending on the location of the object to observe/perceive/discover. This visual scanning modeling could be interfaced with other numerical models of human driver in the future, according to the modular and portable approach used for this development. The visual perception module is connected with others modules of COSMODRIVE, with the aim to support the driver's mental representation elaboration and allow the model progression into the environment. But, in addition to visual thus scanning based on mental representations; the module can also simulate visual attention allocation to an on-board screen. In it possible to use the driver model for investigating driving errors due to a visual distraction while driving.