Efficient calibration techniques for large-scale traffic simulators

• An efficient calibration framework for large-scale traffic simulators is proposed.
• The framework is based on metamodel simulation-based optimization (SO) algorithms.
• The metamodel embeds analytical structural problem-specific information.
• It is used to address a calibration problem for the large-scale Berlin network.
• This approach reduces simulation runtime until convergence by over 80% on average.

Road transportation simulators are increasingly used by transportation stakeholders around the world for the analysis of intricate transportation systems. Model calibration is a crucial prerequisite for transportation simulators to reliably reproduce and predict traffic conditions. This paper considers the calibration of transportation simulators. The methodology is suitable for a broad family of simulators. Its use is illustrated with stochastic and computationally costly simulators. The calibration problem is formulated as a simulation-based optimization (SO) problem. We propose a metamodel approach. The analytical metamodel combines information from the simulator with information from an analytical differentiable and tractable network model that relates the calibration parameters to the simulation-based objective function. The proposed algorithm is validated by considering synthetic experiments on a toy network. It is then used to address a calibration problem with real data for a large-scale network: the Berlin metropolitan network with over 24300 links and 11300 nodes. The performance of the proposed approach is compared to a traditional benchmark method. The proposed approach significantly improves the computational efficiency of the calibration algorithm with an average reduction in simulation runtime until convergence of more than 80%. The results illustrate the scalability of the approach and its suitability for the calibration of large-scale computationally inefficient network simulators.