面向低轨卫星网络的多维资源表征及分配方法OA

Low Earth Orbit(LEO)satellite communication networks,with their multidimensional heterogeneity,high speed node mobility,and time varying link characteristics,pose new challenges for resource allocation.Existing allocation methods focus usually on optimizing a single metric such as throughput or latency,making it difficult to simultaneously satisfy both the Quality of Experience(QoE)for users and the overall system performance.To address this issue,this paper first constructs a multidimensional resource representation model based on a Temporal Graph(TG)within the 3GPP Non Terrestrial Network(NTN)framework.The model incorporates the resource types of LEO satellite networks(spectrum,power,time slots,etc.)and transmission constraints(link reachability,delay,etc.),enabling the characterization of resource demand,supply,and dynamics over time and providing a systematic mathematical description for subsequent optimization algorithms.Then,by using a large set of video service data,a pseudo subjective analysis method is employed to develop a QoE assurance performance evaluation model for video services through the fitting of mathematical models to experimental data.With the objective of maximizing the QoE,a reinforcement learning based resource allocation algorithm,Q Learning based Maximum QoE(QL MQoE),is designed.Experimental results show that,compared with existing allocation algorithms,the QL MQoE can improve the QoE by 7.9％,therefore confirming the effectiveness of the proposed algorithm.