How traffic conditions affect the driver's internal state? A multi-level approach for improving advanced driver assistance systems

7th AHFE - International Applied Human Factors and ergonomic conference 2016, Orlando, ETATS-UNIS, 27-/07/2016 - 31/07/2016; Advanced driver assistance systems (ADAS) are designed to improve driving safety by reducing the risk of human error. Some of these systems process information about the driver's internal state such as stress, fatigue, vigilance or drowsiness in order to assist him and to prevent potential risks (Wang, 2010). Except for few systems, most of ADAS exploit the driver's internal state by means of behavioural or physiological measurements independently of the driving context. A difficulty for innovation in the development of efficient ADAS consists in the fact that the internal state is dynamically constructed through the driving event and affective states felt by the driver (e.g., mental wellness, relaxation or satisfaction) and reciprocally, that the internal states could have an effect on driving behaviour. Nevertheless, to our knowledge, the driver's behaviour merged with physiological data has not yet been taken into account in the assessment of his/her global internal state when considering the driving context. The present study investigated how an external factor such as the traffic conditions and time pressure can influence the driver's behaviour and internal state as a whole. For this purpose, an experiment had been conducted on a driving simulator examining how the drivers performed the same three different driving events (left-turn, overtaking and pedestrian events) under two conditions: baseline and experimental. Traffic density and time pressure were increased in the experimental condition when compared to the baseline condition so as to affect the driver's internal state. The effects of Condition and Event on the driver's internal state were investigated through electrophysiological (ECG and EDA), behavioural (driving performance parameters) and subjective (questionnaires) measures. Each driving event was followed by a car-following task to assess its impact on the following driving behaviour. The results suggested that the drivers react with more accuracy to changes in speed of the lead vehicle after the events in the experimental condition, probably because of an increase in vigilance when also considering the electrophysiological driver's activity. Nevertheless after the sudden occurrence of a pedestrian crossing the road, the drivers lowered other aspects of performance such as standard deviation of lateral position (SDLP). Thus, in absence of an avoidance manoeuvre (e.g., car-following), SDLP could constitute a pertinent behavioural parameter that has already been demonstrated as an indicator of a decrease in performance (Harrison & Fillmore, 2011). Besides, it was observed that the amplitude of the skin conductance level that is very sensitive to stress/mental workload, decreases very quickly after the events. In this regard, some links between physiological and driving data are evidenced. The present study demonstrates the interest to combine electrophysiological and behavioural parameters for assessing overall changes in the driver's internal state. Whereas the interpretations have to be confirmed, some recommendations can already be made for improving ADAS in the automotive industry.