Towards the Optimal Solution for the Routing Problem in Vehicular Delay Tolerant Networks: A Deep Learning Approach

Abstract

Vehicular networks have the goal to provide a communication framework for moving vehicles, road infrastructure and pedestrians. Such kind of networks is the entrance to a new era of services that will make areas such as security and safety, information, transactions, entertainment and sustainability (green transportation) more efficient than they are today, especially in the upcoming era of autonomous vehicles and self-driving cars. However, the severe nature of vehicular environments makes efficient inter-vehicular communication very difficult to achieve. Vehicular Delay-Tolerant Networks (VDTN), as these  etworks are called, have very sparse, intermittent connections, and the lack of a fixed topology gives rise to one of the main challenges that they face: packet routing. A range of routing algorithms has been proposed in recent years to optimize communication in vehicular networks, and significant progress has been made in the matter but the quest for the optimal performance is still ongoing. In this paper, we explore a deep learning approach to the routing problem in these scenarios and propose a routing architecture that helps routers make packet forwarding decisions based on the current conditions of its surroundings. In order to assess the performance of the proposed architecture, simulations were run showing important gains in terms of network overhead and hop count with respect to popular routers, while maintaining acceptable packet delivery rates and average delivery delays.