A new methodology for identification the formation of traffic bottlenecks in real-time

In an era of unprecedented global urbanization, urban society is facing a growing demand for mobility, which creates enormous pressure on the urban road networks [1, 2]. As a result, we face severe traffic congestions and increasing fuel consumption that in the long-run increases air pollution. In the short term, however, the increase in traffic loads, alongside the increase in the use of mobile navigation apps, disconnects the natural linkages between traffic planning and land use allocation.
Recent work presented an innovative approach that uses a percolation process to identify streets that constitute a momentary bottleneck in urban traffic networks [3]. This method not only identifies the bottlenecks but also presents the decomposing of the entire urban network into separated clusters. To identify the temporal clusters, the researchers suggest examining the temporal availability of each street in comparison to its capacity. By doing so, they identify each street at each given moment as blocked or functioning in terms of traffic flow. However, these clusters (defined as strongly connected components) do not guarantee light-traffic flows within them.
In this paper, we present a new method for identification of real-time bottlenecks in a dynamic urban traffic network. It is common to consider a bottleneck as a street where the traffic flow downstream is free while its upstream is congested. Here, we consider downstream and upstream, not just in space, but also in time. Thus, in the presented work, we define a traffic bottleneck as the source of a jammed cluster in terms of both time and space. In other words, a bottleneck is not just any loaded street, but one that affects its upstream traffic flow in a shock-wave effect.
We tested this method on real-time data of the traffic flows in Tel Aviv and in London centers, which was collected from Google Directions API, every 15 minutes over a week. Our findings suggest that by using our method, we can identify, in real time, the formation of traffic bottlenecks. This information can be used not only to evaluate the cost of each bottleneck (in terms of man-hours for example), but also to improve traffic-light control systems by prioritizing the most critical bottlenecks in real time. Additionally, the analysis of the bottlenecks can be carried out at all scales, i.e. local roads or even national ones. Thus, it can be used to solve traffic congestions at different road-networks as it pinpoints the areas that need to be prioritized.