Modeling of decision-making behavior for discretionary lane-changing execution

Modeling of decision-making behavior for discretionary lane-changing execution (DLCE) is fundamental to both movement simulation and controlling design of automatic vehicles. The existing gap acceptance models ingored the nonlinearity of drivers' DLCE decision-making behavior. Therefore, this study tries to analyze and simulate the DLCE decision-making behavior using the real trajectory data. First, a threshold of the lane-changer's lateral velocity is introduced to identify the starting point of DLCE process based on vehicle trajectories from the NGSIM data set. In the following, the empirical analysis based on traffic state variables at the instant of accepting DLCE event are presented, which prove the necessity of modeling DLCE decision-making behavior with machine learning method. Then, we propose a DLCE decision-making model using the Support Vector Machine (SVM). For verifying the prediction performance, the proposed model is compared with the Nagel's model based on the NGSIM data set. The comparison results indicate that the proposed model using SVM outperforms the Nagel's model in predicting the DLCE decision.