An Integrated Bayesian Classification and Gaussian Mixture Method for Signal Phase and Timing Estimation Based on Floating Car Trajectory Data

The information on signal phase and timing (SPaT) is critical for traffic management and signal optimization. However, it is sometimes difficult to obtain SPaT information for traffic engineers in practice. In a connected vehicle environment, floating car trajectory data provides a potential way to derive accurate SPaT information in a convenient manner. To this end, this study proposed an integrated approach for SPaT estimation. Specifically, based on the vehicle trajectory data along the arterial, the naive Bayesian classification was first used to classify non-stopping and stopping trajectories. Then, the Gaussian mixture model was adopted to derive the cycle length and phase timing parameters by exploring the physical queuing process of vehicle movements based on classified trajectories. Last, the proposed method was verified along one arterial in Shenzhen City, China. The results demonstrate that the proposed method enables to achieve desirable estimation accuracy of SPaT parameters.