Measuring driver behavior and driver-vehicle interaction

Lucas P.J.J. Noldus, Ph.D.
Noldus Information Technology BV, Wageningen, The Netherlands

In spite of worldwide research on autonomous driving, cars, trucks and buses will have drivers for the foreseeable future. These drivers need to operate their vehicles in dense traffic amidst an increasing variety of traffic participants, facing a dashboard that is filling up with digital displays and nomadic devices for navigation, communication and entertainment. ADAS systems may be taking over certain driving tasks, but the driver remains ultimately in charge. Cognitive overload, distraction and drowsiness are lurking! Not surprisingly, these are important topics in current driving science. Driver monitoring and measuring driver-vehicle interaction are indispensable methods for understanding, e.g., driving behavior of different age groups, transition of control from car to driver, optimal automation level, how drivers experience different instrument designs, and effect of substances on driving performance.
Driver studies can be performed in driving simulators or in real vehicles. Driving simulators are excellent tools to measure and evaluate task performance of drivers under controlled conditions. They are commonly used in research and development of in-car electronics, dashboards and interior designs. Multiple subjects can be tested with precisely the same vehicle and traffic conditions, and the effects of varying these conditions on operator cognitive load, attention and fatigue can be investigated. Furthermore, in a simulator it is relatively easy to measure detailed information about the driver’s behavior and performance using eye tracking, facial expression and body posture analysis, and physiological sensing (heart rate, skin conductance, brain activity, etc.). Traditionally, car simulators and driver monitoring tools are separate, disconnected systems. However, in order to understand the relationship between specific aspects of the task and the driver’s behavior and performance, it is imperative that all data sources in a multimodal test environment are properly synchronized and integrated. For instance, if the simulator displays an approaching vehicle and the operator looks at that object, the system as a whole does not know that the operator looks at the vehicle. However, by fusing data from the eye tracker (which knows at which screen position the driver is looking) and the simulator (which knows what is being displayed at that screen position), the driver’s visual perception and subsequent actions can be logged in a fully automated manner, ready for an integrated analysis. This multimodal approach also allows us to derive higher-order mental states (e.g. cognitive workload) from the data originating from different sensors.
A lot of progress has been made in recent years in the development of tools that offer a tight integration between driving simulation and automated measurement of the behavior and physiology of the driver at work. With sensors becoming smaller and less obtrusive, they can also be mounted in real vehicles, allowing experiments under naturalistic traffic conditions. These technical trends offer exciting opportunities for researchers in the field of driving science to investigate new concepts, to test product prototypes, and to develop safer vehicles that are easy to operate for all drivers.
Lucas Noldus is the founder and CEO of Noldus Information Technology BV (, a developer of software tools and integrated measurement systems for the study of human behavior and human-system interaction, headquartered in The Netherlands. Numerous automotive companies, universities and research institutes use Noldus tools for studies on driver behavior and driver-vehicle interaction.

Categories: Blog