Graph analysis on ATCS data in road network for congestion detection

Abstract

This research is development a framework for detecting congestion on the urban road network. ATCS (Area Traffic Control System) data in Bandung city with traffic volume are used in congestion detection process. Traffic flow data is collected by vehicles detector located at crossroads within 15 minutes. To compute spatial correlation, graph modelling are used in the adjacency matrix. Assuming the location of the detector as the vertices and the direction of the vehicle as the edge, the graph modeled with vehicle's detector location and the flow direction at nine locations on road nework. The adjacency matrix used consists of 3 matrices in each period of time, which describes the order of spatial distances traveled by vehicle at the intersection location. To calculate spatial correlation, the autocorrelation function and the cross-correlation function which are derived from Pearson's simple correlation is used to looking influence at each location on road network. The result of calculation of spatial correlation, shows the existence of seasonal pattern on the autocorrelation results even though the value scale is getting smaller as it increases time lags. This provides that the process of calculating cross-correlation functions and it can be concluded that the volume of vehicles at each location that are connected in the road network can be known by making observations in the time series of previous seasonal periods. The conclusion that can be formulated that graph modeling is needed to simplify the spatial correlation calculation process by performing the graph representation into a matrix. The Simpson rules on cross-correlation results, can be detected congestion at intersection locations on the road network to find the most critically locations causing congestion on the road network at time periods.