Coupling Urban Road Travel Time and Traffic Status from Vehicle Trajectories by Gaussian Distribution

Urban road travel time estimation is of great importance for intelligent transportation applications, such as the vehicle navigation. With the ubiquitous sensing data, e.g., the mobile sensing of vehicle trajectories, real-time urban road travel time estimation is feasible. Though many road travel time estimation models have been developed by incorporating the road traffic status as an important influencing factor, including the tensor decomposition-based models and neural network-based models, there have been no studies exploring how road travel time is correlated to the corresponding road traffic status. In this paper, we propose to study the coherence of these two traffic indicators by utilizing a probability function. Specifically, we model the road travel time and traffic status with a Gaussian distribution by considering them as two time series signals. Further, we propose to estimate the road travel time with only the road traffic status as observations via the conditional Gaussian distribution. Experiments on realworld datasets demonstrate that road travel time can be estimated with quite a good accuracy by the conditional Gaussian distribution, which further indicates that road travel time and the corresponding traffic status can be captured by a nonparametric Gaussian distribution.

Automated summary

In this paper, the coherence between road travel time and traffic status is studied by utilizing a probability function. The experiments show that road travel time can be accurately estimated using the conditional Gaussian distribution. This research is of great importance for intelligent transportation applications.