Stochastic Calibration of Entry Control Facility Microsimulation Models Using Distribution Matching

Abstract: Entry control facilities (ECFs) control access to restricted areas, typically using guards to verify that drivers have access. For transportation engineers, this provides an operationally complex multivariate environment, with significant costs associated with building or altering an ECF. By definition, efficiency and cost effectiveness can be improved if the checkpoints can be modeled accurately prior to construction. Previous studies have shown that microsimulation models tend to outperform macroscopic models because of the stochastic nature of checkpoints. However, in previous security checkpoint microsimulation studies, the processing rates are assumed to be normally or uniformly distributed, drivers’ lane choices are not calibrated to consider queueing on checkpoint lanes or drivers intended directions at post-checkpoint intersections, and queueing is not considered as a target of calibration. To address these gaps, this study introduces a holistic multi-objective calibration method for ECF traffic microsimulation models. To account for drivers’ lane choices, a dynamic lane choice model considering post-checkpoint driver directions and ECF queue lengths is developed. Non-normally distributed processing times calibration methods are introduced using a four-moment distribution calibration. Additionally, parallel computing strategies are applied to improve the speed of ECF model calibrations. Case studies at three entry control facilities are performed and compared against real-world data to evaluate the methodology. The proposed methodology was able to calibrate each checkpoint to a pareto-optimal solution achieving statistical significance for both processing rates and queueing. Additionally, the proposed parallel processing strategy was able to reduce the calibration time by 39.9%.