A computational model for car drivers situation awareness simulation: Cosmodrive
Résumé
The aim of this research program, so-called COSMODRIVE (for Cognitive Simulation Model of the DRIVEr), is to develop and implement a cognitive simulation model of the car driver able to drive a virtual car in order to dynamically interact and progress into a virtual road environment (so-called SiVIC), through a dual "Perception-Cognition-Action" regulation loop based on both automatic processes (i.e. Skill-based behaviour and decisions) that can be performed in parallel without any attentional effort, and controlled processes (rule and knowledgebased behaviour or decision), requiring cognitive resources to be performed. The central cognitive components supporting this regulation loop in COSMODRIVE are drivers' mental representations of the road environment. These mental models are formulated in working memory (from a matching process between operative knowledge - the "Driving Schemas" and the "Envelope-Zones" - and the perceptive information extracted in the road scene through perception), and they correspond to the driver's Situation Awareness at a given time. Moreover, these mental models are used for anticipation of future situational states (i.e. expectations), in order to make decision an implement driving action allowing the driver to progress on the road and to safely interact with other road users. After having presented the research objectives and the theoretical background supporting this research, we will present some empirical data collected among human drivers in order to investigate visual distraction effects on driving performances and driver's Situational Awareness. Then, the forth section will more particularly present the recent model enhancement focused on Human error simulation, in terms of erroneous SA due to visual distraction while driving. Some empirical data collected among 20 human drivers will be briefly introduced, and a typical case of human errors simulation due to visual scanning of an additional screen will be presented (crash resulting to an erroneous/ not updated SA and a too late detection of a lead car breaking). The conclusion will explore the next steps of this research.