Traffic Allocation Strategy Based on Reinforcement Learning in Two-layer Satellite Network

In MEO-LEO satellite network, a MEO can establish connections with multiple LEOs at the same time. In order to maximize the use of satellite network resources, traffic allocation between MEO and LEOs is very important. This paper first describes the two-layer satellite network system model and point out the importance of traffic allocation among multiple LEO to MEO. Then, by quantifying the traffic resources that can be received by MEO and traffic requirements from multiple LEOs to MEO, and taking into account the remaining visible of MEO to LEOs, a traffic allocation strategy based on Q-learning in two-layer satellite networks was proposed, which aims to maximize the utility of LEOs to MEO in a given time. Then the paper grades the LEO traffic requirements to reduce the computational and storage complexity of the system. Finally, the effectiveness of the proposed strategy is verified by simulation.