Calibration of Micro-Simulation Traffic-Flow Models Considering All Parameters Simultaneously

Abstract

This study proposes a methodology to calibrate simulation-based microscopic traffic flow simulation models. This methodology has the capability to calibrate simultaneously all the calibration parameters as well as demand patterns for any type of network. Global and local parameters were considered as well as performance parameters for driver behavior and vehicles. Demand patterns, in terms of turning volumes, were included in the calibration framework. The primary contribution of this paper is the simultaneous consideration of multiple performance measures, such as link counts and speeds, while calibrating all model parameters across various time periods. This represents a very comprehensive approach for the calibration of simulation-based microscopic traffic flow models. Previous studies claim to consider all model parameters. However, they do not consider multiple performance measures simultaneously. In addition, in the experimental framework, most of them used pre-calibrated parameters and demand patterns. A Simultaneous Perturbation Stochastic Approximation algorithm was used to search for the vector of the model’s parameters that minimizes the difference between actual and simulated network states. The simultaneous consideration of all model parameters and multiple performance measures was motivated by issues associated with convergence and stability. The effects of changing the values of the parameters were taken into consideration to adjust them slightly and simultaneously. This resulted in a small number of evaluations of the objective function. Three networks were calibrated with excellent results. The first network was an arterial network with link counts and speeds used as performance measurements for calibration. The second network included a combination of freeway ramps and arterials, with link counts used as performance measurements. The third network was an arterial network, with time-dependent link counts and speed used as performance measurements. The experimental results illustrate the effectiveness and validity of this proposed methodology. The same set of calibration parameters was used in all experiments.