Reinforcement learning and neuro‐fuzzy GNN‐based vertical handover decision on internet of vehicles

Abstract

SummaryThe Internet of Vehicles (IoV) has become a lot of concern for scientific society because of its rising importance in creating an intelligent transportation system. However, the existing systems addressed handover and network selection difficulties, but they did not solve the necessity for frequent handover problems. Hence, in this research, handover decision difficulties are removed by using a novel state‐action–reward–state‐action reinforcement learning method in which it learns the current policy action to determine whether a handover decision is required in the fast‐moving vehicle therefore the handover failure problem is reduced. Moreover, the existing fuzzy rules do not consider non‐line‐of‐sight and the learning capabilities of the neural networks to select the suitable network therefore it creates network selection difficulty. Hence, the network selection difficulties are removed by using a novel fuzzy‐based graph neural network in which the fuzzy logic is connected to the graph hierarchical structure to select the right network therefore the efficient and suitable network is selected on the internet of vehicle communication. Moreover, the existing process is to select the routes completely established on the ancient broadcast history of the vehicles and their channel metrics which causes the frequent handover and poor performance of the network. Hence, the routing selection problem is eliminated by using a novel cluster‐based routing algorithm in which the vehicle was divided into clusters to select the efficient route of the internet of the vehicle therefore the best routing path is carried out. As a result, the proposed system is improving the quality of the service and reducing the handover failure and delay of the system.