Developing an effective model for identifying congestion influences in different road segments from trajectory data

Abstract

Traffic congestion is one of the most common problems all over the world. They bring about considerable economic loss, increase travel time and huge pollution. Detection of road traffic congestion and its management has been a popular area of research for scientists. The attention is growing to find a solution to critical traffic management in big cities and urban areas. Generally, congestion originated from a road segment propagates through other road segments towards other parts of a road network. Most of the recent researches on congestion detection can only render binary decision about the occurrence of congestion which is unable to address the performance metrics, such as intensity, accuracy, residual effect of congestion on a road segment. Thus mere detecting, measuring a quantitative value of congestion provide more information of real time traffic scenario. Though the propagation of congestion provides some information about the dependencies among the road segments, it cannot measure the influences among them. It motivates to develop an effective model that represents congestion influences among the road segments in a road network. Average road speed is the basic requirement to determine the road traffic congestion and their influences. Therefore, taxi trajectory data is a good option for its availability and reliability. In this work, a novel model has been introduced to measure the congestion and identify their influences among road segments in a road network. The model generates a mathematical expression for measuring the congestion considering the effect of change in average speed with respect to current average speed and the time decay of congestion to get the residual effect of congestion in time series. The model selects connected road segment and implements simple linear regression to measure the influence value. We construct a graph for visual representation of the model where the nodes represent the congestion value of each road segment and the edges represent the congestion influence value between connected road segments. Finally, we validated our model by experiments on a large real-time taxi trajectory data in an urban road network.