智能反射表面的主被动联合波束成形分组模型优化算法

Optimization Algorithm of Joint Active-passive Beamforming Grouping Model on Intelligent Reflecting Surface

针对智能反射面(IntelligentReflectingSurface,IRS)联合主动和被动波束成形优化中计算复杂度过高的问题,该文基于经典的半正定松弛算法和交替优化算法提出了一种对本地IRS进行分组的新模型,通过降低每个IRS信道矩阵的秩来降低优化算法的复杂度。具体的,首先通过已知的IRS相位得出BS(Base Station)的次优发送波束,其次通过BS的发送波束计算得出分组后各IRS面的次优相位,经过不断交替迭代上述两步优化得出最优发送波束的结果。仿真结果表明,合适的IRS分组数量、分组方式与交替优化迭代次数可以在几乎不影响原算法性能的情况下,大大减少交替优化算法的复杂度。

The paper proposes a new model for grouping local Intelligent Reflecting Surfaces(IRS)to reduce the computational complexity of joint active and passive beamforming optimization.The model is based on classical semi-definite relaxation algorithms and alternating optimization algo⁃rithms.Specifically,the paper first obtains the suboptimal transmission beam of the Base Station(BS)through the known IRS phase,and then calculates the suboptimal phase of each IRS surface af⁃ter grouping through the BS transmission beam.The optimal transmission beam is obtained by itera⁃tively optimizing the above two steps.Simulation results show that an appropriate number of IRS groups,grouping methods,and iteration times can greatly reduce the computational complexity of the alternating optimization algorithm without significantly affecting its performance.