Predicting the driver's turn intentions at urban intersections using context-based indicators

Many improvements have been made in the area of vehicle safety and pedestrian protection; however, urban intersections are still black spots for Advanced Driver Assistance Systems (ADAS). One of the main reasons is the uncertainty of the future driving direction at intersections. Due to that uncertainty, the early activation of an intersection ADAS will lead to high false positive rates, while, in contrast, a late activation of an intersection ADAS will lead to a low accident-reduction potential. This tradeoff is described as warning dilemma. In order to solve the warning dilemma, an approach to predicting the driver's turn intention at urban intersections is introduced. The novelty of the approach is its context-based prediction of the future driving maneuver several seconds before the driving trajectory changes. To predict maneuvers, indicators are used to encode the context information together with vehicle data. A system setup including prediction results of the system is described.