延迟CSI反馈下的协作NOMA系统用户选择方法

A User Selection Scheme for Cooperative NOMA System with Delayed CSI Feedback

为了提升多用户协作非正交多址接入(Non-orthogonal multiple access,NOMA)网络在信道状态信息(Channel state information,CSI)反馈延迟下的性能,提出一种近用户选择策略,目标是在满足所选近用户服务质量约束的同时最大化远用户的信息传输速率。分析了反馈延迟影响下,网络基于所提近用户选择策略的中断概率和分集增益。为了降低CSI反馈延迟影响,进一步设计了一种卷积神经网络与循环神经网络结合的深度学习架构对过时CSI进行预测。理论分析表明,反馈延迟影响下,多用户选择分集增益消失,系统分集增益为1。仿真结果表明,所提深度学习架构能够显著提升反馈CSI准确性,有效降低系统中断概率。基于预测CSI的近用户选择方法提升了该协作NOMA网络的系统中断性能。

In order to improve the performance of the multi-user cooperative non-orthogonal multiple access(NOMA)network with delayed channel state information(CSI)feedback,this paper,firstly,proposed a near-user selection strategy,which maximizes the data rate of the far-user while satisfying the service requirement of the selected near-user.The outage probability and the diversity gain of the NOMA network based on the user selection strategy were analyzed with the influence of delayed CSI feedback.Then,this paper designed a convolutional and recurrent neural networks-based deep learning architecture to predict outdated CSI.The theoretical analysis shows that the diversity gain is 1 with the impact of delayed CSI feedback,which means the diversity gain of multi-user disappears.The simulation results show that the proposed deep learning architecture can improve the accuracy of CSI and the system outage probability can be significantly reduced.The near-user selection scheme based on predictive CSI can improve the system outage performance of the cooperative NOMA network.