Can Road Traffic Volume Information Improve Partitioning for Distributed SUMO?

Microscopic vehicular simulations can be computationally intensive due to the sheer size of the road network and number of vehicles. One solution is to parallelize the simulation through distribution and concurrent execution of the scenario being simulated. To enable distributed simulation of an area, the partitioning of the map into different areas for parallel execution on different nodes is required. How the map is partitioned is also a critical factor for distributed simulation, as a poor partitioning can lead to a communication overhead and/or an imbalance of workload among the computing nodes. In this paper, we ask: Can traffic volume information improve the classical structural partitioning algorithms? In the context of improving distributed simulation in SUMO, we propose a modification to three existing mechanisms for road network partitioning, SParTSim, Smart Quadtrees and Quadtrees, with the aim of creating more balanced partitions (in terms of workload) derived from traffic volume data.