Investigation of Traffic Congestion and Safety Indices Using Trajectory Data Collected with Drones

The use of drones offers an innovative traffic data collection framework that allows to capture detailed and accurate vehicle trajectories. However, deploying them for an entire day would be prohibitively expensive due to the limitations of their batteries and current regulations. In this paper, we show how commercial, low-cost drones can be used to monitor traffic during a full day in an efficient way throughout the day. More specifically, we conducted a traffic data collection campaign using a swarm of drones in several intersections in the congested city center of Manchester, UK, where traffic streams were recorded during both the morning and afternoon peak hours. Using advanced computer vision techniques, we extracted vehicle trajectories that enabled us to conduct an in-depth safety analysis of two busy signalized intersections, the second one also including an elevated freeway. Our analysis conducts a comprehensive traffic safety assessment, identifying the areas of the intersections that are prone to crashes. We utilize well-known surrogate safety measurements (SSMs) like the Time-To-Collision (TTC) and the Post-Encroachment-Time (PET) to detect risky interactions and observe their spatial distribution. Furthermore, we investigate the causes of these risky interactions through the identification of conflicting critical movements. Finally, we use the data to establish a relationship between traffic variables like speed with the frequency of near crash events.