面向天地融合网络多模态中继回程链路选择方法OA

To address the problem of differentiated Quality of Service(QoS)requirements for terminal services and heterogeneous backhaul link resource adaptation in Space-Terrestrial Integrated Network(STIN),a Reinforcement Learning(RL)-based intelligent selection method for multi-modal relay backhaul links is proposed.In this method,a Delay-Sensitive(DS)service graded queue man-agement mechanism is designed to achieve adaptive resource scheduling in dynamic network environments.Based on this,a multi-ob-jective optimization problem for the packet loss rate of DS services and the throughput of Best-Effort(BE)services is formulated,em-ploying deep RL algorithms to implement online optimization of backhaul links.Simulation results show that,compared with Fixed Pri-ority(FP)and Weighted Fair Queuing(WFQ)strategies,the proposed method effectively reduces the packet loss rate of DS services and simultaneously improves the throughput of BE services in high-load scenarios.