基于机器学习梯度下降法的OAM对准算法

OAM Alignment Algorithm based on Machine Learning Gradient Descent Method

OAM为通信系统增加了一个新的维度即“模分复用”,有望借此增大信道容量。但是由于涡旋电磁波的特殊性,OAM对收发天线的轴心对准要求很高,轴偏差对于信道容量的影响很大。我们以能量谱方差为损失函数,提出了一种基于机器学习梯度下降法的高效对准算法,并进行仿真。仿真结果表明了这种算法可以在损失函数未知的情况下,根据当前位置测量得到的数据调整参数,以很快的收敛速度实现参数的最优化,使OAM系统收发机精确快速对准。相对于已有的对准算法,这种算法更适合于机器实现,而且不依赖于初始状态。

OAM adds a new dimension to the communication system,namely“module division multiplexes”,which is expected to increase channel capacity.However,due to the particularity of eddy electromagnetic wave,the OAM has high requirements on the axis alignment of the transmitting and receiving antennas,and the axial deviation has a great influence on the channel capacity.With energy spectrum variance as the loss function,an efficient alignment algorithm based on machine learning gradient descent method is proposed and simulated.The simulation results indicate that the proposed algorithm can adjust the parameters according to the data obtained from the current position measurement when the loss function is unknown,and optimize the parameters with fast convergence speed,so that the transceiver of OAM system can be aligned accurately and quickly.Compared with the existing alignment algorithm,this algorithm is more suitable for machine implementation and does not depend on the initial state.