Performance Evaluation of Traffic Congestion Detection Algorithms in Real-Life Scenarios

Traffic congestion in urban cities has substantial economic and social effects. Roads in urban cities have become more and more crowded. However, it is challenging to upgrade the cities' infrastructure and open new roads for traffic. There-fore, Intelligent Transportation Systems (ITS) introduce Artificial Intelligence (AI) based solutions to help keep the traffic flow in a free state. Identifying the traffic congestions in real-time is critical in ITS solutions as it can provide time to prevent the congestions' transitions through the city road network. In our previous work, we have proposed three novel algorithms to detect the congestions in real-time [1], [2]. The algorithms were verified using synthetic traffic data. In this paper, their performance evaluation using real-life data from the state of California [3] is introduced. The experimental results show that the algorithms are capable to be utilized in real-life scenarios. Our algorithms overperformed the other methods from the literature in terms of detection rate and false alarm rate. Moreover, they have achieved the best performance in terms of detection time by identifying the congestions faster than the other algorithms, which is crucial for the city traffic operators to intervene on time to avoid the transition of the congestion to other roads' sections.