摘要：AbstractIn a recent paper (Chavan et al. (2016)), the authors proposed a dynamic model of traffic behaviour on (Indian) roads, where lane discipline is not maintained even when the roads are wide and multiple vehicles can travel abreast. The primary hypothesis in this model was that each driver is influenced by an assortment of neighbouring cars and not only by the vehicle immediately in front. The proposed equations claim to be a model for driver/vehicle movement on such roads and hence should be able to predict actual motion of real vehicles. In this paper we experimentally verify this claim, after suitable modification of the theory to make it applicable for heterogeneous traffic. Videos of normal traffic on a sample road in Mumbai city, India, is recorded from the top of an adjacent high-rise building. Detailed motion information of groups of cars is extracted through novel image processing techniques. The proposed model is initialized with the extracted data and the computed trajectories are compared with the actual ones calculated from the images. It is verified that the proposed model can accurately predict complex maneuvers, such as overtaking, sideways movements (lane-changing) and avoiding collisions with slower moving vehicles.
关键词：Keywordslane-less trafficcar following modelmulti agent systemimage processing
