How do Young Drivers Behave in Intersection Dilemma Zone? Comparison of Drivers’ Stop-go Intention Recognition Model

Young drivers are overrepresented in road crashes, half of which occur at intersections in urban areas. One of the main contributing factors in intersection crashes is the presence of dilemma zone. The main motive of this study is to investigate factors influencing young drivers’ driving performance and predict drivers’ stop/go intention based on their driving performance. A driving simulator experiment was conducted under three kinds of traffic densities (low, medium and high), two types of intersections (X-intersection and T-intersection) and two kinds of yellow duration (3 s vs. 4 s) conditions. Both drivers’ eye-movement data and vehicle control data were collected to explore drivers driving performance when approaching intersection. Behavioral measures such as braking initiation to stop and eye-movement measures such as drivers’ fixation time, saccade time were found significantly different under different traffic densities, intersection types and yellow duration conditions. Both CNN-LSTM model and BP-HMM model were then built with significant measures as input data to predict drivers’ stop-go intention. Model performance evaluation indicated that CNN-LSTM model and BP-HMM model both have excellent performance in solving the dilemma decision prediction problem, and the prediction accuracy reached 94.97% and 97.79%, respectively.