Computational Simulation of Visual Distraction Effects on Car Drivers’ Situation Awareness

This paper presents a computational modeling approach for negative effects simulation of visual distraction while driving a car. In order to investigate these effects, an experiment was firstly implemented on a driving simulator. Twenty participants were invited to perform a car following task in different driving conditions (12 driving scenarios), with or without a secondary task of visual distraction. Empirical data collected through this experiment show that visual distraction negatively impacts the driving performance at both perceptive and behavioral levels, and then increase the risk of having a crash. Beyond these effects on the observable performance, the aim of this study is also to investigate and simulate these distractive effects on mental models of the road environment. Indeed, driver’s decisions and behaviors are based on a temporal-spatial mental model, corresponding to the driver’s situation awareness (SA). This mental representation must be permanently updated by perceptive information extracted from the road scene to be efficient. In case of visual distraction requiring off-road scanning, mental model updating is imperfectly done and driver’s actions are thus based on a mental representation that can dramatically differ from the situational reality, in case of a critical change in the traffic conditions (e.g. sudden braking of the lead car). From these empirical results, a computational model (named COSMODRIVE for COgnitive Simulation MOdel of the DRIVEr) was implemented for simulating visual distraction effects and human errors risks at perceptive (visual scanning changes) cognitive (erroneous Situation Awareness) and behavioral levels (late reaction time and crash risk increasing).