A NGSA-II Based Parameter Calibration Algorithm for Traffic Microsimulation Model

Abstract

Traffic microsimulation is a widely used tool in the field of traffic engineering, this tool will help engineers to evaluate the current traffic system and analyze the alternate plans. In the process of applying traffic microsimulation, the step of model calibration is critical and difficult. The main task of this step is to adjust the parametric values of simulation model, and this task can localize the microsimulation model. Model calibration process includes 3 main sub-steps, and one of these sub-steps, searching for the optimal parametric values, is confusing and time-consuming. This parameters searching work is always handled by experienced traffic engineers, and associated with a great amount of statistic analysis. In order to improve the efficiency and accuracy of this step, a type of methods, which takes parametric values searching problem as nonlinear optimization problem, has been discussed by several papers. However, there are still some disadvantages in those methods. In this paper, we converted the parametric values searching problem into a multi-objective optimization problem, and applied NGSA-II, a fast and elitist multi-objective genetic algorithm, to solve it. In addition, this approach is applied to calibrate a microsimulation model for a freeway merging section as an example.