Compared with the embedded car-following model in AIMSUN, the new car-following model is 20% better in terms of errors reduction [20]. 3. Significance of the Research Many severe crashes occur at signalized intersections today due to signal violations and so it is important to study buy Pazopanib the red-light running prevention at intersections. Compared with the other traffic segments, the driver behaviors close to signalized intersections are more difficult to represent by models due to the random individual vehicle’s decision, intensive interactions between vehicles, and the complex feedback mechanism between

vehicles and the traffic signal system. The driver behaviors study at signalized intersections will help identify the possible reasons of certain unsafe vehicle maneuvers, such as the RLR, and eventually help develop countermeasures to mitigate the safety hazards. In the past, it is commonly assumed that the RLR is caused by the dilemma zone

and therefore most of the related research of the red-light running issue focused on how to minimize the vehicles in the dilemma zone. However, some recent research on vehicle trajectories during yellow and all-red clearance reveals that vehicles may still run red lights at low speeds in which the dilemma zone issue hardly exists. This finding implies that there must be other factors than the dilemma zone to contribute to the red-light running. Intuitively, drivers at congested intersections may be more likely to take the RLR risk to cross the intersection in order to avoid further waiting. Or the drivers may be just distracted and fail to observe the traffic lights. Obviously, the reasons for RLR are complex and difficult to be represented with the traditional modeling method. Meanwhile, although it is difficult to precisely analyze the reasons for the individual RLRs, the red-light runners may still

share some common kinematic patterns, such as shorter headways from their leading vehicles or faster speeds at yellow onsets. These kinematic features can be retrieved from the vehicle trajectory data collected via the radar, video imaging detectors, or the connected vehicle technology in the future. In this paper, the authors explored various ANN networks to approximate the driver behaviors Entinostat during the yellow and all-red clearances. The inputs of ANNs (i.e., vehicles’ kinematic features) were captured and calculated based on the vehicles’ trajectories during the yellow and all-red clearance. The well trained ANN model then served as the fundamental predictive model to identify the possible red-light runners. The collision avoidance measures were activated then to avoid potential crashes. 4. Methodology 4.1. Problem Representation 4.1.1. ANN Model Inputs It is assumed that a potential RLR event begins at the yellow onset when a driver has to decide whether to cross or stop according to his safety perception. The safety perception is psychological and determined by many factors.