下行RIS-NOMA的用户集群方法

User clustering method for downlink RIS-NOMA

为提高用户分布非均匀的场景下可重构智能表面(RIS)辅助下行非正交多址(NOMA)通信系统的性能,提出了一种分布式RIS辅助下行NOMA的用户聚类方案。首先,采用自适应几何分布(AGD)聚类算法划分用户集群(UCs),从而为各UC匹配RIS。然后,利用分式规划(FP)方法将非凸最大化频谱效率问题转换为凸优化问题。最后,为UC内用户逐级进行功率分配(PA)与被动波束形成(PB)。仿真结果表明,与谱聚类(SPC)、K均值(K-means)、高斯混合模型(GMM)聚类方案和正交多址(OMA)方案相比,在AGD方案下,功率对频谱效率提升了7%、14%、19%和42%的增益,反射面对频谱效率提升了16%、19%、26%和40%的增益。

In order to enhance the performance of downlink non-orthogonal multiple access(NOMA)communication systems assisted by reconfigurable intelligent surfaces(RIS)in scenarios with non-uniform user distribution,a distributed user clustering scheme for downlink NOMA assisted by RIS has been proposed.Firstly,the adaptive geometric distribution(AGD)clustering algorithm is utilized to cluster users into user clusters(UCs)and match RIS with each UC.Then,the fractional programming(FP)method is employed to transform non-convex maximization and rate problems into convex optimization problems.Finally,power allocation(PA)and passive beamforming(PB)are performed sequentially for users within UCs.Simulation results demonstrate that,compared to spectral clustering(SPC),K-means,Gaussian mixture model(GMM)clustering schemes and orthogonal multiple access(OMA),the AGD scheme yields approximately 7%,14%,19%and 42%gains in powerto-spectral efficiency,while RIS contributes approximately 16%,19%,26%and 40%improvements in spectral efficiency.