Analysis of Three Phase Traffic Flow Theory under Pre-Crash Condition

Abstract

Road crashes are a major concern for people of different backgrounds. Like other professionals, transportation engineers have given higher priority to this. Earlier crash prevention involved the identification of crash hotspots and then implementation of safety measures like providing appropriate speed limits, improving physical conditions, etc. Another aspect of crash prevention is the development of crash prediction models which help in evaluating the crash risk of future roads as well as identify hotspots alongside their underlying relationships with road geometry, aggregated traffic characteristics, weather and environment, etc. Statistical or artificial intelligence based methods are used to identify the likelihood of crashes in these models. These models can be static, i.e., using largely aggregated traffic data, or real-time data, which uses traffic data for short time window. Apart from predicting crash probability, the real-time crash prediction models have also been used to understand crash mechanism. These models separate the traffic conditions into normal traffic which is associated with traffic conditions when crashes do not take place, and pre-crash conditions, i.e., traffic conditions just before a crash. Some studies have evaluated two phase traffic flow theory with pre-crash traffic data. However, no investigations were done in order to investigate the variations of three phase traffic flow theory under pre-crash conditions. This research investigates how variables of the three phase theory behave under pre-crash conditions. This study specifically focuses on the wide moving jam phase. It has been previously established that the downstream jam velocity of the wide moving jam is within 10km/h to 18 km/h. Shibuya-3 of the Tokyo Metropolitan Expressway is chosen as study area for its dense and uniform spatial distribution of detectors. The data has been used to estimate the downstream jam velocity of both normal and pre-crash conditions. The detector based method and correlation based methods are used for velocity estimation. The downstream jam velocities under normal conditions are within the range of 10 km/h to 18 km/h. The downstream jam velocities are also within the range of 10 km/h to 18 km/h. However, there are significant variations in between pre-crash and normal conditions. It was also vi found that the detector based and correlation based methods yielded contradicting results when consistency of the results was compared.