Prediction of Driver’s Risky Behaviors Based on Vehicle Trajectory Data at Signalized Intersections

Abstract:

A trajectory prediction model and a driver’s hidden decisionmaking process prediction model were established based on the collection and analysis of vehicle trajectory data at intersection influenced areas. Moreover, based on these two models, the rules for recognizing driver’s risky behaviors were built. The three together constitute the overall driver’s risky behaviors prediction method at signalized intersections. It is found that the overall accuracy of this prediction method is 88.89%. This prediction method realizes vehicle’s individual dynamic trajectory prediction, utilizing driver’s hidden decisionmaking process as an important variable in prediction of driver’s risky behaviors.