부산대학교 도서관

Traffic counts (or link counts) are defined as cumulative traffic in the lanes between two consecutive intersections on a road network. Established methods of link count estimation assume the availability of count data at a set of basis links: a minimum subset of all the links of a network that still allow complete network traffic count estimation. If traffic count data are missing even at some basis links, current research must introduce additional assumptions, on path flow or historical data, to compensate. In this research, we present an approach to estimate the missing basis link count without the need for historical data or path-flow information, thereby overcoming the limitations of state-of-the-art estimation approaches. We develop a stochastic method using a canonical correlation analysis-based constrained minimization problem for estimation purposes. The proposed method has been validated with real-world link count data collected in Melbourne, Australia, between 2016 to 2019. The validation results indicate that we can achieve an accuracy of up to 90% in the real world, despite the unknown traffic patterns of the estimation period. Depending on the time of day, the modelling strategy selected, and the consistency of input data available at the road intersection, the estimation accuracy varies. The proposed methodology is useful when there is a general shortage of data since there is inadequate infrastructure for data collection in many major cities around the world.