Calibration of Wiedemann 74 Car Following Modeling using Departure Headways

Abstract: Calibration refers to the adjustments of model parameters to best replicate the real conditions of a transportation system (Federal Highway Administration (FHWA), Traffic Analysis Toolbox Volume III, 2019). While calibration is only one step in the microscopic traffic simulation process, it is vital to ensure that the model replicates the system. There are numerous guides available from FWHA and State Departments of Transportation (DOTs) that provide general guidelines about which parameter to change and their value envelopes. However, this high-level guidance does not typically provide a methodology on how to calibrate models. Within the literature, there are several calibration approaches that use genetic and other algorithms to find the optimum parameter sets. While these approaches may find an “ideal” parameter set, their black-box nature does not provide transportation professionals any transparency. Aside from that, these models can risk overfitting. Additionally, the measures of effectiveness (MoEs) utilized for calibration often do not capture the direct effect of the car following parameters. Rather, MoEs such as travel time capture the secondary aggregate impacts of multiple modeling aspects in addition to car following, such as lane changing, signal settings, desired speeds, etc., making it difficult to isolate the impact of car following parameters. For instance, corridor capacity is often driven by intersection control, in many cases traffic signals, in which car following is critical. Car following parameters directly impact the throughput of traffic signals, more specifically, the saturation headway. While acceleration, lane changing, etc., do have an impact on broader MoEs, the most significant impacts of the calibration parameters occur at the intersection.
Given the gaps in practice and in research, the current effort develops an approach that is (1) practical to implement, and (2) flexible to allow judgment from an experienced modeler. Calibration of the Wiedemann 74 model involves (1) calibrating the standstill distance (ax), and (2) calibrating the additive and multiplicative part of safety distance (bx). Additionally, this effort considers calibration of the desired speed distribution. Demonstrating the relationships between the different parts of the Wiedemann model allows for a flexible approach. Calibration of the Wiedemann parameters focuses on the departure headway distribution. It is seen how the ax and bx parameters influence the saturation headway and subsequently the throughput and capacity of the facility. The desired speed is calibrated with a similar approach, leveraging the underlying speed distributions in the model and the field. Combining calibration of the car following parameters and desired speeds ensures that critical aspects of driver behavior are replicated in a simulated environment. Modeling efforts that better reflect driving behaviors can eventually help decision-makers make more informed decisions.