Model calibration to simulate driving recommendations for traffic flow optimization in oversaturated city traffic

Long queues at signals cause fuel-consuming stop-and-go traffic. Empirical measurements have shown that driving behaviour can have an important impact on queue length and thus on the occurrence of stop and go traffic. This led to the question of whether even a few vehicles can have a measurable influence on the traffic situation in congested city traffic. In this work we use a complete microscopic spatiotemporal measurement of congested city traffic at a signal to i) calibrate a both longitudinal and latitudinal driving model and then to ii) examine how changes in single vehicle's driving behaviour could improve the situation. The model calibration is realized using a genetic algorithm. In this way, a realistic heterogeneous traffic scenario that has similar properties as empirical traffic could be simulated. We then show that already changing the behaviour of a single vehicle per traffic light cycle can significantly reduce the number of vehicles waiting in queues.