An improved spatial process model for mining urban traffic state information from macroscopic view

Abstract

Traffic state is an important indicator and usually used for describing road network performance. Traditional traffic theory modeled traffic performance from fluid dynamics and time-series analysis. However, these models cannot obtain satisfactory result under real traffic situatuion, espcially macroscopic environment. From priori knowledge, within a discrete time interval, many of the vehicles that traverse one road link would traverse neighbor road links as well. Thus it is reasonable to think that traffic state of urban road network has spatial association. Therefore, firstly this paper verifies spatial correlation of urban traffic state using spatial statistics theory and represents the validity of study, then based on the stationary temporal nature of urban traffic state during typical traffic time periods, uses spatial process model to describe it in different time periods. The study is tested on Nanchang's urban road network with sparse road link travel speeds derived from approximately 1,200 floating cars (GPS-enabled taxis). The experiment results show that spatial process model is reasonable and practical to describe complex urban traffic state from macroscopic view during fixed time period and especially, it is conceptually simple and thus, easy to achieve in practice. Therefore, this study can contribute to mine pontienal traffic information from spatial perspective and provide a new research idea for traffic state analysis and other relative traffic studies.