Dynamic routing planning method for large-scale low-orbit satellite networks based on location guided and multi-agent DQN network

Large-scale low-orbit (LEO) Satellite Networks have the characteristics of wide coverage and low delay, and have attracted a lot of attention. However, due to the fast moving speed of LEO satellites, the topology of LEO networks changes frequently. In order to improve the utilization of network resources and the speed of routing calculation, this paper proposes a dynamic routing method for large-scale low-orbit satellite networks based on multi-agent DQN location guided networks. With the training of a large amount of prior data, the proposed method can enable the network nodes to make routing decisions based only on the surrounding environment. In addition, the transmission domain partition scheme is proposed, which can accelerate DQN convergence by reducing the routing scope and decreasing the satellite nodes during training. As the traffic distribution of satellite networks is not uniform in reality, a queuing model based on population density distribution is established. The simulation results demonstrate that this method has better performance than the existing methods in terms of packet loss rate, and model convergence speed and can decrease the end-to-end latency.