Spatial-temporal adaptive network partitioning for urban traffic signal control

Abstract

In response to rapidly growing and diversifying traffic demand, it is necessary to develop a network partitioning method that could achieve real-time global optimal performance and adapt to traffic network evolution. In this paper, an adaptive partitioning method is presented, which achieves optimal partitions at runtime and determines appropriate Time-of-Day breakpoints to update partition results simultaneously. For each time interval, partitioning schemes are firstly assessed in terms of modularity by taking roadway geometry, real-time traffic flow information, and signal timing into account. Two values are attained from the assessment: the maximum modularity of the optimal partition and the modularity obtained from the existing partition. Then the existing partition is updated, provided that the relative deviation of these two values exceeds a given threshold for a certain number of successive time intervals. Experimental results show that the above-mentioned partitioning scheme outperforms some notable traffic control techniques in modularity in the spatial aspect. In the temporal aspect, the updating scheme can well respond to varying traffic conditions and yield significantly higher average modularity.