A Metamodel Simulation-based Optimization Approach for the Efficient Calibration of Stochastic Traffic Simulators

This work considers the calibration of stochastic microscopic traffic simulators. It formulates the calibration problem as a simulation-based optimization (SO) problem, and uses metamodel SO ideas. The main idea is to embed within the calibration algorithm an analytical problem-specific description of how the calibration parameters are related to the simulation-based objective function. Preliminary results on a toy network are presented. Ongoing work applies these ideas to the calibration of a Berlin metropolitan area network.