Estimating traffic flow at urban intersections using low occupancy floating vehicle data

Fixed-section detection methods, with radar and video as representatives, frequently encounter incomplete detection data at controlled intersections because of high construction costs and insufficient maintenance. This results in ineffective signal control strategies. On the other hand, mobile detection methods, represented by floating cars, can perceive both macro and micro spatial-temporal characteristics of traffic flow. However, their current low penetration rate limits their ability to provide sufficient data support for signal control at intersections.To address this issue, this paper proposes an innovative method to obtain more accurate flow rates for each phase at an intersection through simulation approximation of calibrated parameters. This method utilizes the Webster delay theory to quantitatively describe the relationship between phase flow and vehicle delay, allowing for the inverse estimation of flow rates. These estimated flow rates are then refined using the proposed Radial Basis Function (RBF) neural network approximation method to achieve higher accuracy. Comprehensive experimental results demonstrate that the proposed method effectively improves the accuracy of inverse flow data estimation. This enables the effective utilization of low-penetration-rate floating car data (FCD) in signal control at urban intersections. By leveraging this innovative approach, signal control systems can make more informed decisions, leading to smoother traffic flow and improved traffic management in urban areas.