Implementation of a Driver Aware Vehicle Using Multimodal Information

There has been recent interest in an integrated approach to driver safety which focuses on the overlapping and interacting area of the role of driver, vehicle and road environment in driving safety. Many active safety systems such as adaptive cruise control, parking assistance and lane keeping systems have been developed to target these intersecting regions. However, dynamic driver state was not properly taken into account in the safety systems because the selection of dominant attributes and modeling architectures for state detection are not fully established yet. In order to categorize driver state in terms of driver wellness, researchers at MIT suggested a modified inverted-U shaped curve which depicts the relationship between arousal level and driving performance. This paper demonstrates an implementation of a driver aware vehicle platform to detect driver state based on the MIT wellness concept. In order to detect driver state, various overt and covert measures such as driving performance, visual attention, physiological arousal and traffic situation should be collected and interpreted. The main focus of this paper is to provide implementation techniques for the synchronized data collection and integration of inputs from multiple domains.