Reinforcement Learning-Based Routing Protocols for Port Internet of Vehicles: Enhancing Biotechnological Supply Chain Logistics

Abstract

Efficient and reliable logistics are essential for the commercialization of biotechnological products, especially in port environments where the transportation of temperature-sensitive and high-value biotech goods demands stable and secure communication networks. To address the challenges of metal occlusion, signal interference, and highly dynamic operational tasks that often disrupt communication in port-based Internet of Vehicles (IoV) systems, this study proposes a Reinforcement Learning-Based Port Internet of Vehicles Routing Protocol (PVRP) designed to optimize biotechnological supply chain logistics. The proposed PVRP employs an adaptive learning routing strategy utilizing historical traffic flow data and is structured into two key components: (1) Road Section Selection, achieved through multi-dimensional Q-table generation using Q-learning to identify the optimal next-hop road section for data relaying; and (2) Vehicle Selection, which determines the optimal relay vehicle within the selected road section. A port-like simulation environment was developed using SUMO simulation software to evaluate the protocol's performance. Comparative analysis against advanced routing protocols such as GPSR and QGriD demonstrates that PVRP improves the packet delivery success rate by 25.6% and 22.1%, respectively, while reducing communication delays by 19.5% and 13.1%. These results highlight the protocol’s superiority in enhancing packet transmission rates and minimizing communication delays, making it highly applicable for supporting biotechnological logistics in smart port scenarios. By ensuring reliable and efficient transportation networks, PVRP contributes to the secure and timely delivery of biotechnological products, thus facilitating robust commercialization processes within the biotechnology industry.