Trajectory Data-Driven Pattern Recognition of Congestion Propagation in Road Networks

The congestion pattern recognition in urban road networks helps for recognizing the bottleneck in road networks and assisting to route planning. With the widespread use of GPS devises in vehicles, it is possible for researchers to monitor the traffic condition of urban transport networks at a road level. In this paper, we utilize the trajectory data of vehicle GPS to detect the road travel speed by matching points of trajectories to road segments. A fuzzy clustering based method is proposed to classify the road congestion level according to the road traffic conditions. Further, the road network is clustered by the proposed snake clustering algorithm, so that the road network is divided into congested and uncongested areas. This paper studies the congestion propagation problem and propose to employ the dynamic Bayesian network for modeling the congestion propagation process. Taking the real road network of Shanghai and the dataset of GPS trajectories generated by more than 10,000 taxis, we evaluate the pattern recognition based congestion prediction method. It shows that the proposed model outperforms the competing baselines.