Routing Strategy in LEO Satellite Networks: A Multi-Agent Reinforcement Learning Approach

Low earth orbit (LEO) satellite network is promising to meet the low latency, extensive coverage, and flexible networking requirements for future 6G communications. To realize efficient data transmission in dynamic LEO satellite network, an appropriate routing strategy is essential. In this paper, we investigate a multi-task LEO satellite network. Specifically, the routing strategies for all tasks are designed to maximize the quality of service (QoS) for data transmission. Since the formulated problem is a mixed-integer nonlinear programming problem, a multi-agent reinforcement learning-based routing algorithm (MARL-RA) is proposed to address this problem. Simulation results demonstrate that the proposed algorithm significantly improves the QoS of data transmission and reduces the end-to-end (E2E) delay by 50%, 33% and 16.7% compared to other benchmark schemes, respectively.