The 20th and 21st joint COTA International Conference of Transportation Professionals (CICTP 2020-21)

Key: Data Fusion; Speed estimation; Neural Network; High resolution; Microscopic Simulation Obtaining high resolution traffic states of large-scale freeway is always a significant topic for both transportation engineers and researchers. This paper presents a machine learning based high resolution speed estimation method for large-scale freeway using two data sources. Two low resolution heterogeneous traffic data are collected from microscopic simulations with different error distributions. A neural network based model is implemented fusing the two data sources and improving both time and space resolution of traffic estimations. The validation results and the sensitivity analysis indicate that the proposed method is feasible and suitable for large-scale freeway speed estimation. The performance of the model is acceptable, and the model could indeed improve both time and space resolutions of the estimations. Author List, Yanjie Gui, 502250970@qq.com, School of Transportation, Southeast University, Fan Ding, dyinfan1129@gmail.com, School of Transportation, Southeast University Hanxuan Dong, 493296757@qq.com, School of Transportation, Southeast University Jiankun Peng, pengjk87@gmail.com, School of Transportation, Southeast University Huachun Tan, tanhc@seu.edu.cn, School of Transportation, Southeast University

CICTP