Routing in VANETs Based on Q-learning and received signal strength information

Abstract

Vehicular Ad hoc Networks (VANETs) present a difficult challenge for both vehicular and infrastructure routing. This study focuses on addressing this challenge, particularly in old Eurocentric urban environments, by introducing a routing decision criterion based on Ad-hoc On-Demand Distance Vector (AODV) protocol and enhanced by Q-learning. The reinforcement learning approach is mainly empowered by the received signal strength indication between each exchange, which aims to aid in optimal route prediction in such environments. Additionally, this research introduces novel Q-learning transitional reward factors that occur during the route request and reply stages, as well as in Hello packet exchanges. The performance of the proposed algorithm is evaluated on OMNeT++ and SUMO. Simulation results demonstrate comparative packet delivery ratios and end-to-end delay efficiencies, with a notable performance edge in scenarios involving common obstructions such as buildings.