Edge Intelligence Enabled Data Transmission in IoV: Integrating Link Optimization and Packet Routing

In the Internet of Vehicles (IoV), effective data transmission between vehicles enhances awareness of potential risks and traffic anomalies, significantly improving traffic safety and efficiency. However, in complex urban environments, existing routing protocols face challenges such as severe attenuation of wireless signals, inefficiency of the three-handshake mechanism, and low reusability of multi-hop links, These issues drastically affect the performance of routing protocols. To address these problems, this paper develops an edge intelligence-enabled routing protocol (EIRP), integrating link optimization and packet routing. The proposed EIRP consists of three key components. Firstly, we propose a traffic awareness-based multi-hop link construction scheme to construct and maintain links among interactions according to traffic conditions. The primary goal of this scheme is to establish stable links before message transmission to improve its efficiency. Secondly, we design a multi-hop link scheduling scheme to enhance data transmission performance and improve the utility of the constructed multi-hop links. Thirdly, we propose a cooperative multi-agent reinforcement learning (MARL)-based packet routing scheme, leveraging individual rewards to address the sparse rewards challenge. Extensive experiments are conducted to evaluate the performance of the proposed EIRP. The results demonstrate that EIRP significantly improves the packet delivery rate and ensures low communication delay in various urban scenarios.