Driver Assistance Systems: Intervene, Warn, or do nothing?

Kurzfassung

The current discussion about future advanced driver assistance systems (ADAS) focuses on sensor development and sensor data fusion and resulting new options to prevent collisions. Assistance functions are being developed technologically driven or to solve special isolated cases of accidents, e.g. blind spot monitoring, head-on collisions etc. Only recently new, more integrated concepts dealing with driver assistance are being discussed. On the one hand, ADAS are supposed to support the driver when he or she is either overburdened or - in the other extreme - falling asleep. On the other hand, ADAS are designed to prevent or correct errors of the driver which might result in an accident. The first approach has a strong focus on driving comfort while the second is centred on increasing safety. To the first goal, information is needed about the driver’s state and a strategy has to be developed how to deal with levels of extreme workload or with falling asleep. To the second goal, ADAS have to evaluate the driver’s actions with regard to their probability to result in an accident. Furthermore, an intervention strategy has to be developed in order to prevent the accident. A combined concept is presented which was developed within the context of the EU project SPARC (Secure Propulsion using Advanced Redundant Control, s. www.sparc-eu.net) which is about an electronic co-pilot in trucks and cars. Test drives from 15 subjects on German Autobahn, highways and cities were recorded with the DLR ViewCar to develop an algorithm to evaluate the driver’s activity with regard to being active, quiet or not active. A virtual co-pilot analyses the driver’s behaviour and evaluates if it becomes dangerous. An HMI strategy combines this analysis of the driver’s state with the criticality of the situation to suggest the best ADAS intervention in the current situation. Basically, if the driver is active, the ADAS intervenes only when the situation becomes critical. If the driver is quiet or not active, the ADAS warns relatively early and intervenes if the driver does not react. The paper presents the algorithm developed, the HMI strategy and first steps to implement and assess this strategy by drivers.