Developing a fast and effective network free-speed calibration procedure for agent-based transport simulations

Agent-based simulations play an essential role in transport planning. To ensure realistic and reliable outcomes, an accurate network model is essential. In this study, we propose a fast and effective network free-speed calibration procedure for agent-based transport simulation frameworks. To evaluate and quantify its performance, we implement the procedure in an agent-based transport simu?lation framework, MATSim. The results demonstrate that the proposed procedure effectively aligns the free-speed of the network with reference data. Moreover, it outperforms existing network calibration methods in output quality and computational efficiency. Specifically, the new approach further reduces the error statistics related to free-speed travel time by over 50% compared to existing methods. Additional experiments also reveal that the new network calibration procedure can still deliver good results even with limited reference data.