Safety evaluation for driving behaviors under bidirectional looking context

ABSTRACT Under the traffic environment of the Internet of Vehicles, especially with the development of vehicle-to-vehicle technologies (i.e., V2V), drivers not only frequently perceive the vehicle information for preceding vehicles but also can actively check the real-time information of following vehicles with the assistance of V2V technologies. However, those traditional traffic flow models where driving behaviors are driven only by the preceding vehicle information are no longer suited to describing this situation. Therefore, a bidirectional cellular automaton model is proposed under active bidirectional looking (ABDL) context, which involves the actively obtained succeeding vehicle information and includes two special cases (i.e., the forward-looking only (FLO) and backward-looking only (BLO)). Meanwhile, a methodology of giving the sensitivity coefficients in the randomization probability of braking is introduced, which helps categorize drivers into three different sensitivity types (i.e., aggressive, normal, and conservative). Subsequently, various kinds of homogeneous and heterogeneous vehicle fleets are generated depending on the combinations of stimulation type (i.e., ABDL, FLO, and BLO) and the sensitivity types, whose safety performance can be evaluated by the average danger coefficients. Finally, through numerical experiments, it is concluded that considering both the preceding and succeeding vehicle information under V2V while driving is conducive to improving the safety performance, and the vehicular flow state, stimulation type, sensitivity type, and orders between different stimulation or sensitivity types have an important effect on the safety performance.