An Evaluation of Space and Graph-partitioning Methods for Distributed Road Network Simulations

We evaluate three different approaches to road network partitioning for simulation. The first approach consists of algorithms that rely on dividing the road physical network spatially into rectangular blocks of different sizes. Our second approach uses a graph representation of the road network, and uses meta-heuristic search algorithms to partition the graph. The final hybrid scheme builds clusters of nodes in the graph representation of the road network, based on spatial information. One of the most important contribution in this paper is the novel formulations of the well-known meta-heuristic approaches of simulated annealing and genetic algorithms to solve the road network partitioning problem. We also present a detailed experimental comparison of the three major approaches, using the road networks of several world cities (Table 1).