In this paper, we propose a new high-performance computing approach to road trafc simulation. Multi-agent road trafc simulators are the most accurate and realistic that currently exist, however they are very resource intensive and the data are massive when simulating a metropolis or a region. The main contribution of this paper is the presentation of a new concept based on the Unite and Conquer approach, allowing to set up several parallel executions with different parameters. Applied on MATSim, one of the most present multi-agent trafc simulators in the literature, it allows to reduce the number of iterations needed to converge to the optimal solution. In this paper, we show how the Unite and Conquer approach applied to MATSim brings a substantial gain in computing time and points out its potential application to other multi-agent simulators.

No abstract