Probabilistic Prediction of Driving Behavior on Country Roads

Driving behavior prediction plays an increasingly important role in the development of autonomous driving functions and Advanced Driver Assistance Systems (ADAS). The more precise a prediction is, and the longer the temporal horizon it covers, the easier it becomes to correctly assess a situation and effectively perform a driving or intervention task. In most cases, this is relevant above all for maintaining safety. While on highways and in cities, it is primarily the surrounding traffic that influences the behavior of vehicles, on country roads it is often necessary for a driver to adapt the driving behavior to the road topology, making it challenging for behavior prediction. Therefore, we build a prediction model that uses the characteristics of the road to estimate the future range where vehicles will be located on country roads. The magnitude of the influence of different factors on the prediction is investigated, and the concept is validated in terms of accuracy and conservativeness. We show in simulation that the method can improve both comfort and fuel economy for an Adaptive Cruise Controller (ACC) compared to other prediction approaches.