A network partitioning methodology for distributed traffic management applications

This paper presents a multi-way network partitioning methodology for distributed traffic management applications. The methodology can be used to partition a typical urban transportation network such that: (a) the inter-flow among the resulting subnetworks is minimised; (b) the subnetworks are balanced in terms of their sizes/flow activities; and (c) each subnetwork is connected. Two heuristics are presented. The first adopts a recursive iterative procedure to determine the network's sparsest cuts that maintain the balance and connectivity requirements. The second heuristic adopts a greedy network coarsening technique to determine the most flow-independent subnetworks. The solution quality of these two heuristics is evaluated using hypothetical and real networks with different configurations. The results show that the heuristics can obtain near-optimal solution in significantly shorter execution times.