Discrepancy of Lane Flow Distribution and Lane-Wise Speed Distribution When Simulating Lane-Drop Bottleneck in Simulation Of Urban Mobility (Sumo): A Comparison of Two Car-Following Models

In this paper, Simulation of Urban Mobility (SUMO) is used to investigate the effectiveness of two different car-following models at simulating a lane-drop bottleneck. The Krauss model (the default for SUMO) overestimates the bottleneck capacity by 46% and cannot replicate the lane flow distribution (LFD) and lane-wise speed distribution (LWSD) effectively. In addition, the Krauss model exhibits platooning on the free-flowing lane (the lane that does not need to merge at the bottleneck) when traffic increases to over 3000 vehicles per hour (vph), a pattern not observed in the real world. Wiedemann 99 model, nevertheless, is relatively more accurate at replicating LFD and LWSD on all lanes other than the free-flowing lane, where excessive speed and flow are recorded. The result shows that the Wiedemann 99 model is more appropriate when simulating lane-drop bottlenecks; however, further calibrations of its parameters will be needed to correctly replicate the LFD and LWSD.