A Novel Multi-Objective Routing Scheme based on Cooperative Multi-Agent Reinforcement Learning for Metaverse Services in Fixed 6G

Abstract

The 6th Generation Fixed networks (F6G) with holographic communication and omni-directional sensory coverage is expected to arrive in 2030. Due to the characteristics of cross-integration between the physical and digital worlds, metaverse has been widely recognized as an important application in F6G to be utilized in all walks of life in the future. However, the metaverse applications will generate diversified communication services with differentiated Quality of Service (QoS) requirements, which will be a great challenge for F6G to develop End-to-End (E2E) customized transmission strategies. Traditional single metric-based routing algorithms cannot efficiently orchestrate the network resources to meet the diversified QoS requirements. To solve the above problems, we propose a Cooperative Multi-Agent Reinforcement Learning (Co-MARL) routing algorithm, which measures the differentiated QoS demands through a generic utility function to facilitate multiple agents to solve the multi-objective optimization problem. The simulation results show our scheme outperforms the traditional routing algorithm in meeting the diversified QoS requirements.