High-Resolution Queue Profile Estimation at Signalized Intersections Based on Extended Kalman Filter
Abstract
Modern urban traffic control systems often require high resolution of queue profile (or queue length) information as input for optimization. With the development of communication technolo- gies, such as 5G, connected vehicles (CV) technology is increasingly receiving more attention as an alternative for collecting traffic data. This study proposes a second-by-second queue profile methodology combining extended Kalman filter with shockwave theory utilizing traffic data col- lected from CVs. The proposed method does not require any prior information about the arrival rate of traffic flow and market penetration rate (MPR) of CV. Moreover, most of the previous re- search only utilize the trajectory data provided directly by CVs, while in this paper, information on the traffic environment surrounding a CV is also utilized. The proposed model is tested with sim- ulation data and real-world data from the pNEUMA dataset (a unique vehicles trajectory dataset obtained from the first experiment using a large number of drones), both for under-saturated and over-saturated conditions. To evaluate the estimation results in term of the queue profile, we calcu- late the MAE and RMSE of the estimated queue tail location and queue length. The results show that the method can provide sufficiently accurate results under different levels of MPR of CVs (e.g., when MPR of CV is 5%, MAE is 14.24 m and RMSE is 19.78 m for queue tail location; while MAE is 12.99 m and RMSE is 17.39 m for queue length). Further, the influence of GPS noise in CV data is analysed.
Qishen Zhou
PhD Student
Junyi Li
PhD Graduate
