基于非正交多址和卷积神经网络的光接入网

Optical Access Network Based on Non-Orthogonal Multiple Access and Convolutional Neural Network

针对下一代光接入网,提出了一种基于功率域非正交多址的扩容无源光网络方案。传统的功率域非正交多址网络基于信道估计函数,采用串行干扰消除(SIC)算法进行解调和恢复,需要精确的信道估计,存在差错传播效应。针对该问题,提出了基于修正卷积神经网络(CNN)的信号接收方案,利用大量数据独立拟合各用户信道函数,打破了SIC解调算法中用户间的依赖链,提升了系统的传输性能和公平性。结果表明,在扩容功率域非正交多址无源光网络中,相比传统SIC解调算法,基于修正CNN的接收方案对远端用户(传输60 km)和近端用户(传输20 km)的传输性能可分别提升约0.5 dB和1.7 dB,且其公平性指数更接近1。

This paper proposes a long reach passive optical network based on power-domain non-orthogonal multiple access for the next generation optical access network.Based on channel estimation function,the traditional power-domain non-orthogonal multiple access networks employ successive interference cancellation(SIC)algorithm for signal demodulation and recovery,which require accurate channel estimation and has error propagation problem.To overcome this problem,a signal receiving scheme based on modified convolutional neural network(CNN)is proposed,which exploits a large amount of data to independently fit the channel function of each user,thus to break the dependence chain between users in the SIC demodulation algorithm,and improve transmission performance and fairness of the system.The results show that,compared with the traditional SIC demodulation algorithm in the long reach power-domain nonorthogonal multiple access passive optical network,the receiving scheme based on the modified CNN can improve the transmission performance of far-end users(transmission 60 km)and near-end users(transmission 20 km)by about 0.5 dB and 1.7 dB,respectively,and its fairness index closer to 1.