ITSC 2025 Paper Abstract

Paper WE-EA-T1.2

Shigenaka, Shusuke (National Institute of Advanced Industrial Science and Technology), Nishida, Ryo (National Institute of Advanced Industrial Science and Technology), Yamazaki, Keisuke (National Institute of Advanced Industrial Science and Technology), Onishi, Masaki (National Institute of Advanced Industrial Science and Technology)

Rapid Parameter Calibration Method using Model-Bridge for Microscopic Traffic Simulations

Scheduled for presentation during the Regular Session "S01b-Data-Driven Simulation and Modeling for Smart Mobility Systems" (WE-EA-T1), Wednesday, November 19, 2025, 13:50−14:10, Southport 1

2025 IEEE 28th International Conference on Intelligent Transportation Systems (ITSC), November 18-21, 2025, Gold Coast, Australia

This information is tentative and subject to change. Compiled on October 19, 2025

Keywords AI, Machine Learning Techniques for Traffic Demand Forecasting, Digital Twin Modeling for ITS Infrastructure and Traffic Simulation, Model-based Validation of Traffic Flow Prediction Algorithms

Abstract

This study addresses the offline calibration problem of input parameters in microscopic traffic simulations, with a specific focus on the origin–destination (OD) demand and signal control parameters to minimize the arrival time error at bus stops. These parameters have traditionally been calibrated by repeatedly running computationally intensive microscopic simulations and comparing the results with observed data, resulting in a time-consuming calibration process. Recent studies have introduced metamodels, which are computationally efficient surrogate models that approximate high-cost traffic simulations, and metamodel-based parameter calibration has been reported to be effective for OD demand calibration. However, conventional calibration methods mainly target OD demand, whereas the signal control calibration remains insufficiently explored. Moreover, existing metamodels are not adequately designed to incorporate signal control, potentially leading to unsuitable parameter estimates for microscopic simulations. Thus, we propose a rapid parameter calibration method using a model-bridge that learns the relationship between the metamodel and microscopic traffic simulation. The proposed method is applied to a large-scale urban network in Japan and evaluated through a comparative analysis of 100 synthetic scenarios. The experimental results demonstrated that the proposed method with the model-bridge reduces the calibration time by a factor of 25 compared with the conventional approach.