Due to the fact that the visual perception of drivers for the surrounding environment is usually subject to various constraints and their perception characteristics have significant heterogeneity, lane-changing behavior is often more or less accompanied by collision risk. Especially for intelligent vehicle drivers, precise risk recognition before a lane change has more practical significance than these post-assessment or instantaneous recognition methods. In order to identify the potential collision risk of lane change more accurate, a new method of spatial risk estimation (SRE) considering vehicle trajectory prediction is proposed in this paper. Unlike the existing researches, which are mainly interesting in how fast or accurate to identify the instantaneous or overall risk during lane change, this paper presents a mobile coordinate system to show the relative positions of vehicles clearly. Moreover, 1444 groups of lane-changing samples are established by using the trajectory data provided by NGSIM. Finally, based on the real-time trajectories of the surrounding vehicles, potential conflict events in lane-changing behavior can be classified efficiently by the proposed model, while the location of potential lane-changing conflict points will be visualized by the dynamic coordinate system. The experimental results show that the proposed model can effectively identify potential conflict events during the lane change. Compared with other researches, the recognition accuracy of the model in this paper has a slight improvement, which means that it has more advantages in recognizing the lane-changing conflict events. The findings can provide theoretical support for the algorithm design of an intelligent lane-changing warning system.

CICTP