Distributed multi-cluster dynamic Q-routing for large size traffic grids

Abstract

The field of autonomous vehicles has seen increasing interest over the past decade, giving birth to the next generation of intelligent vehicles. These vehicles are getting closer to performing a multitude of driving-related tasks without any human intervention. However, to achieve full autonomy, a fast and reliable routing strategy must exist to ensure optimal path calculation. Yet, given the large size of modern cities and daily traffic volume, most static and centralised algorithms have huge limitations and will not solve congestion problems. Therefore, in this article, we will propose an architecture that exploits machine learning power, especially Q-routing, coupled with network clustering techniques, to offer a distributed routing solution in a real-size network by partitioning the traffic grid into a manageable size. Furthermore, we will present simulation results that conclude that the proposed architecture offers an effective routing solution in significantly faster computation time.