Sag is a road section with significant vertical gradient change from downhill to uphill, which is a capacity bottleneck that easily causes traffic flow breakdown. To understand the trigger of breakdown, a typical sag section in Xiangyin Tunnel, Shanghai, is investigated using loop detector data in this paper. The empirical analysis indicates that the breakdown is mainly caused by the car-following behavior. Thus, an improved car-following model is proposed to explore the cause of breakdown by incorporating the gravity factor and several human factors (perception error and the asymmetry between acceleration and deceleration process). Combining a simple but effective lane-changing model, numerical simulations are conducted to analyze the influence of these car-following factors on traffic capacity. The results show that the proposed car-following model can describe the traffic flow characteristics of sag section, and the asymmetry is found as the critical internal cause of sag breakdown.

CICTP