GPS data Mining at Signalized Intersections for Congestion Charging

Nowadays more private car trips have caused worse congestion due to the Covid-19 pandemic in many cities. Congestion charging is one of the taxes that is levied on vehicle owners to reduce urban traffic congestion. One of the most important reasons congestion charging is not accepted by the public is the high cost. Monitoring the state of traffic congestion in real time requires a lot of expensive installations. The purpose of this paper is to make congestion charging more accurate and acceptable using artificial intelligent algorithm. Massive real-time Global Positioning System (GPS) data provides new data for road congestion charging. The queuing length at intersections is an important measurement for the degree of traffic congestion, and it is also the basis for road congestion pricing. GPS positioning cannot provide sufficient position accuracy for lane identification of vehicles. In this study, a comprehensive model consisting of a real-time lane identification model and a real-time queue length estimation model is developed based on the traffic shockwave theory using GPS data. The comprehensive model can identify the lane where the queuing vehicle is located and estimate the real-time queue length of the lane. The proposed models were evaluated using field-collected data in Guangzhou, China. The testing results show that the proposed comprehensive model can identify lanes and estimate queue lengths with satisfactory accuracy. The model proposed in this paper provides real-time data for road dynamic pricing in a cost-effective way, which can promote the implementation of congestion charging in cities.