Exploring bus tracking data to characterize urban traffic congestion

Quantification of traffic dynamics is a valuable tool for city planning and management. Metrics such as the vehicle average speed, travel time, delays, and count of stops, can be used to characterize mobility and traffic congestion in an area. However, effective study of mobility data is often hindered by the difficulty of gathering mobility data in a practical, inexpensive, and prompt way.

In this work, we explore the use of city buses as mobility probes, using the existing smart city infrastructure deployed in Aveiro, Portugal. We propose a method for traffic congestion detection considering the low vehicle speed, low traffic flow and road occupancy close to its capacity. Three degrees of congestion are identified using the k-means approach; DBSCAN is used to characterize the typical level of congestion in a road. Using four-weeks of mobility data, it was possible to assess the congestion along the day and for the different days of the week; some road segments proved to be consistently prone to congestion. We also studied parameters of driving safety, considering speed and acceleration.

In this work, we show that knowledge discovery can be applied to mobility data being collected by tracking buses, exploring data that is often collected for other purposes also to characterize traffic congestion. These methods can inform decision makers and are easily ported to other cities.