Digital twin-based integrated estimation, prediction, and control for expressway traffic: an application demonstration

Abstract

This paper presents a digital twin (DT) framework for integrated operation management and control of expressway networks, demonstrated through simulation on a multi-segment corridor with on-ramp merging. The framework follows a closed-loop estimation-prediction-control architecture comprising an Extended Kalman Filter (EKF) for real-time traffic state estimation from sparse sensors, a short-term prediction module based on a macroscopic traffic flow model, and a Model Predictive Control (MPC) optimizer that determines ramp metering rates and variable speed limits. The physical expressway is modeled with the METANET model, which serves as ground truth. Three capability levels are compared, namely no DT, partial DT with EKF estimation only, and full DT with estimation, prediction, and MPC, under free-flow, congestion-forming, and heavy-congestion scenarios. The full DT reduces total travel time by up to 4.2 percent under heavy congestion relative to reactive control, while the EKF maintains density RMSE below 1.2 veh/km/lane and speed RMSE below 1.8 km/h. A sensor-coverage analysis shows that performance is preserved with only 33 percent sensor deployment, indicating substantial potential for infrastructure cost reduction. The framework offers a practical foundation for deploying digital twin technology in real-world expressway operation management.