Maps of Road User Occupancy in Intersections and their Impact on Target Tracking Performance

In future intelligent transportation systems, a variety of detection and tracking systems will interact cooperatively to make transitions through road crossings safer, cleaner, and more efficient for all road users alike. The demand of robust situational awareness is especially high in areas with a high diversity of road users and their interactions, i.e. at intersections and their immediate neighborhoods. Common reliable non-linear tracking methods used for this task, like particle filters, often show a high demand in computational effort. To ease this demand we take a look at intersection geometries to find areas where show a certain level of similarity. To this aim we apply the Information Bottleneck Method to discretize the road intersection area into few groups of cells that can show similar behavior. We then use a particle filter to track a sample bicyclist and use the estimated velocity map to analyze the required computational complexity.