大规模MIMO系统中的改进CG迭代算法

Improved CG iterative algorithm in massive MIMO systems

大规模多入多出技术以其高的频谱效率和能量效率,成为未来移动系统的关键技术之一。但是,随着用户天线数目的增加,多入多出检测算法的计算复杂度会急剧增加,导致在实际系统中难以快速、有效地实现。针对大规模多入多出检测计算复杂度高、收敛速度慢等问题,提出了一种改进的共轭梯度迭代算法。首先将理查森迭代算法的初始矩阵作为牛顿迭代算法的初始矩阵进行迭代;随后将迭代结果作为共轭梯度迭代算法的迭代初始矩阵进一步迭代,在保持较低计算复杂度的同时,实现快速收敛;最后通过理论定量分析了算法的计算复杂度。通过仿真实验对比,分析了多种大规模多入多出检测算法在误码率性能和收敛速度方面的性能。结果表明,相比其他算法,所提算法具有更低的计算复杂度和更快的收敛速度。在调制方式为64QAM、天线规模为32×256或64×1024时,仅需要3次迭代,检测性能即可接近最小均方误差算法。

The massive multiple input multiple output technology becomes one of the key technologies of mobile systems at present and even in the future due to its high spectrum efficiency and energy efficiency.However,as the number of user antennas increases,the complexity of the detection algorithm will increase sharply,which makes the algorithm difficult to implement quickly and effectively in the actual system.An improved conjugate gradient iterative algorithm is proposed to solve the problems of high computational complexity and slow convergence.Under the condition of channel hardening,in this algorithm the initial matrix of the Richardson iterative algorithm is used as the initial matrix of the Newton iterative algorithm for one iteration,and the result of the Newton iterative algorithm is used as the initial matrix of the Conjugate Gradient iterative algorithm for further iteration.Through these steps,the algorithm not only keeps the computational complexity low,but also accelerates the convergence speed.The computational complexity of the algorithm is quantitatively analyzed by theory,and the bit error rate performance and convergence speed of the algorithm are compared with other typical detection algorithms through simulation experiments.Simulation results show that the proposed algorithm has a lower computational complexity and faster convergence speed compared with other schemes.When the modulation mode is 64 QAM and the antenna size is 32×256 or 64×1024,the detection performance is close to the minimum mean square error algorithm only after 3 iterations.