A. Kamel, H. Hozayen, H. Talaat

pdf icon

Pages: 217-234

This paper presents a real-time traffic flow estimation model that uses open-source travel time data provided by Google Cloud Platform (GCP) in addition to physical/operational attributes of road segments. The inclusion of such attributes was targeted to enhance the accuracy of the estimation process, especially in traffic environments with low level of organization and loose adherence to traffic right-of-way. The adopted case study is the Ring Road of Greater Cairo Region, Egypt. Four attributes were selected based on their observed impact on the expressway speed-flow relationship: number of lanes; alignment curvature; transit stops; and prevailing maneuvers. A data collection plan was executed to obtain GCP-based travel speed data for different road segments with concurrent traffic counts. Data was used to train an ANN model that uses GCP-based travel speeds and road segments’ attributes to provide online traffic flow estimates. The developed model reported MAPE of 4.87% compared to 9.41% as best result obtained from a parametric traffic flow model and 6.8% for an extended version of a locally calibrated 3-regime model. Residual analysis reported random distribution of residual errors with 73% of points with residual error between +/-100 pcu/hr/lane. Trained ANN model was used to estimate flow adjustment factors that compare flow estimates at given speeds for different types of road segments to that of a base case one. Estimated adjustment factors ranged from 0.828 to 1.415. Reported results highlight the potential of using open-source GCP travel time data for real-time traffic flow estimation and the significance of including physical/operational attributes of road segments in enhancing estimation accuracy.
Keywords: Artificial Neural Network; ANN; Traffic State Estimation; Google Cloud Platform; Google API

Issues per Year