A Novel Way of Optimizing Headlight Distributions Based on Real-Life Traffic and Eye-Tracking Data Part 2: Analysis of Real-World Traffic Environments Data in Germany

In order to find optimized headlight distributions based on real traffic data, a three-step approach has been chosen. Since the complete investigations are too extensive to fit into one single publication, this paper is the second of three papers. Over the course of these papers, a novel way to optimize automotive light distributions based on real-life traffic and eye-tracking data is presented. Over all three papers, 119 test subjects participated in the studies, with over 15,000 km of driving, including recordings of gaze behavior, light data, detection distances and other objects in traffic. In the first paper, an ideal headlight distribution for straight roads with no other road users was identified. The second paper aims to collect the data required to modify this idealized headlight distribution for use on real roads. The first step is to find the extent to which real roads differ from an ideal, straight road. To do this, the German traffic space was analyzed. A new test vehicle recorded video and GPS data over a selected route. The video data were then evaluated by a machine learning algorithm. Object recognition software was used to find different traffic participants and road signs. Camera calibrations were used to find the exact angles of these objects. Using publicly available road data combined with the recorded GPS data, the video data were split into different road categories, and traffic object distributions were calculated for urban roads, country roads and motorways. The resulting analyses provided representative distributions of vehicles and highway signs along different types of roadways and roadway geometries. The GPS data were also used to find the curvature distributions along the selected route. These data were then used to optimize segment sizes for an adaptive driving beam. Overall, increasing the number of segments above 100 did not have appreciable benefits. These data will also be used in the third paper, where along the same route, the gaze distribution of drivers was recorded and analyzed.