MAEON: An Efficient Weather-Aware Ocean Network Routing Scheme based on Multi-Agent Reinforcement Learning

Shu Yang, Yaofeng Liu, Laizhong Cui, Runsu Zhu

Abstract: Ocean network communication is more and more important nowadays. However, it faces significant challenges due to its heterogeneity, low reliability, and narrow bandwidth. Compared with traditional networks, the problem worsens under severe weather conditions because communication channels, e.g., microwave links, may degrade badly due to weather changes. While many ocean applications put higher QoS (Quality of Service) requirements on communications, it is difficult to meet them due to fluctuating weather changes. Thus, it is more challenging to model and operate the ocean network as more factors may influence the performance.Currently, artificial intelligence opens up new possibilities to meet the challenges because it can adapt to the dynamic changes of the network. In this paper, we formulate the problem by establishing a model between network performance and weather conditions. We try to optimize network utility given the traffic matrix and important weather factors, such as rain, atmospheric absorption, and clouds. To solve the problem, we propose a two-stage and multi-agent optimization algorithm named MAEON (Multi-Agent Efficient Ocean Network). We conduct a comprehensive simulation using generated network topology traffic and real-world weather datasets. We also carry out a case study with datasets from a real-world scenario. The results show that MAEON can improve performance by 21.5% compared with traditional algorithms.