Abstract
Keywords - workload, driving manoeuvres, dynamic estimation, advanced driver assistance systems
Information about the current driver workload is seen as crucial for new advanced driver assistance and information systems. Besides the specific functionalities these systems also have the general aim of optimizing drivers´ workload in order to reduce critical situations, errors and discomfort. As sensory equipment of modern series-production vehicles is limited and hardly allows to measure conventional workload indicators as e.g. heart rate, steering wheel reversal rates and subjective measures to a sufficient extent, an alternative approach to provide information about the actual driver workload was developed. Two main issues have been in focus: the dynamic and continuous information about external stress factors and resulting internal driver strain as well as likewise dynamic online access to this information within the vehicle.
Within the driving context workload depends on different external stress factors, e.g. driving manoeuvres and environmental conditions (road characteristics and traffic density). As these stress factors could change in a highly dynamic way this in turn demands a comparable dynamic assessment of driver strain. Within several empirical studies we therefore developed and tested a continuously subjective rating method for experimental settings in order to measure the effect of external stress factors on driver strain multimodal. Using the empirical results a .model of driver workload is build up.
To achieve the second issue of dynamic online access to driver workload within series-production cars an alternative approach is build up. By directly measuring external stress factors via nowadays already available sensors information within vehicles, resulting driver strain is indirectly estimated online using the workload model mentioned above. As the empirical studies resulted in strong effects of driving manoeuvres and local traffic densities on driver workload an online recognition of these factors (via CAN-Bus and laser scanner) is used for measuring stress factors directly. Combining the information about current stress factors and the effect sizes of the workload model leads to an estimation of actual driver strain. Further steps contain the implementation of both the online manoeuvre and local traffic density recognition and the detailed workload model within our driving simulator for validation of the model.