Detection of leader–follower combinations frequently observed in mixed traffic with weak lane-discipline

Abstract

We quantitatively clarified that, in an actual mixed traffic, the distributions of the types of leaders and followers for each vehicle type have biases toward the distributions predicted based on the number of each type of vehicle. Vehicular traffic in some countries is heterogeneous, having weak lane-discipline, i.e., 2D mixed traffic. Previous studies have clarified that the order of vehicle types in a vehicle platoon influences the traffic properties. It has been further suggested that a certain vehicle type, e.g., motorcycles, exhibits herding or grouping behavior. However, such grouping phenomena have not been quantitatively investigated. In this study, we aim to extract the vehicle combinations frequently observed in an actual 2D mixed traffic. By utilizing video footage that was shot overlooking a road in Mumbai, India, the extended vehicle order, i.e., leader–follower network, was detected first. Then, we compared the observed distribution of leader–follower combinations and the estimated distribution based on expectations calculated using the number of each type of vehicle, which revealed discrepancies between these two distributions in the leader–follower network. For example, motorcycles more frequently became leaders and followers of themselves than estimated. Moreover, they less frequently had leader–follower relationships with other types of vehicles. Specifically, auto-rickshaws and motorcycles are mutually exclusive. Based on these obtained results, two groups appear to exist: a group of motorcycles and another group consisting of auto-rickshaws, cars, and heavy vehicles. Our results implied that segregation phenomena comprises “isolation,” as well as filtering, gathering, and dispersion, as proposed in a previous study. The results also showed that not only size but even agility (i.e., quickness in longitudinal acceleration or deceleration) and speed play different roles for each type of vehicle in the respective segregation process, though the cause of the mutual exclusiveness between motorcycles and auto-rickshaws is still unclear. From the asymmetry of the frequency of leader–follower relationships, we proposed two types of preference, i.e., “potentially desiring” and “unintentionally attracted.”