Freeway Merging Decision Analysis on the Acceleration Lane

Numerous traffic accidents and near-crashes are caused by lane-changing maneuvers. A merging vehicle (MV) on a freeway acceleration lane (AL) has to merge into the faster-traveling mainline stream before the end of the AL. With the space constraint and speed difference pressure, freeway AL merging is more complicated and riskier compared with discretionary lane changes. Adequately characterizing the driving behaviors in different merging scenarios and in different sections of the AL can help understand the mechanism of the merging process and benefit merging decision assistance technology development. To resolve this issue, a data acquisition platform was established to acquire natural merging process data. Statistical analysis results demonstrated that the existence of the leading vehicle on the mainline had no significant effect on the merge position of the MV. The average acceleration or speed of the MV in the “merge” events was relatively larger than that in the “wait” events. As the MV neared the end of the AL, the risk it could bear increased, and the acceptable lag gap threshold and time-to-collision threshold decreased. The main challenge of a merging decision model lies in how to balance the risk acceptance behavior and the spatial constraint of the AL. Based on the analysis, an adaptive boosting (Adaboost) model was established to model the merging decision of the MV considering the current and historical data of the environment and the effect of AL. The overall accuracy of the model turned out to be 98.9%.