New Gateway Selection Algorithm Based on Multi-Objective Integer Programming and Reinforcement Learning

Abstract

Connecting vehicles to the infrastructure and benefiting from the services provided by the network is one of the main objectives to increase safety and provide well-being for passengers. Providing such services requires finding suitable gateways to connect the source vehicles to the infrastructure. The major feature of using gateways is to decrease the load of the network infrastructure resources so that each gateway is responsible for a group of vehicles. Unfortunately, the implementation of this goal is facing many challenges, including the highly dynamic topology of VANETs, which causes network instability, and the deployment of applications with high bandwidth demand that can cause network congestion, particularly in urban areas with a high-density vehicle. This work introduces a novel gateway selection algorithm for vehicular networks in urban areas, consisting of two phases. The first phase identifies the best gateways among the deployed vehicles using multi-objective integer programming. While in the second phase, reinforcement learning is employed to select a suitable gateway for any vehicular node in need to access the VANET infrastructure. The proposed model is evaluated and compared to other existing solutions. The obtained results show the efficiency of the proposed system in identifying and selecting the gateways.