基于机器学习的毫米波大规模MIMO混合预编码技术

Machine Learning-based mmWave Massive MIMO Hybrid Precoding

毫米波大规模多输入多输出技术是提高5G移动通信容量的核心技术之一,其中混合预编码技术作为大规模MIMO系统中最关键的技术而被广泛研究。采用传统的迭代算法解决混合预编码问题通常导致较高的计算复杂度和严重的系统性能损失。机器学习方法由于其具有自适应学习和决策的优势而被应用于混合预编码器的设计工作中。在机器学习的基础理论上提出了一种采用交叉熵优化策略的混合预编码算法,通过迭代更新具有稳健误差的交叉熵损失函数得到最佳的混合预编码器组合,该组合被证明可以实现理想的传输总和速率,可以显著提高系统的能量效率。

Millimeter wave(mmWave)massive multi-input and multi-output(MIMO)is one of the key technologies to improve the capacity of 5G mobile communications,where hybrid precoding has been widely studied as the most critical problem in massive MIMO systems.The traditional iterative algorithms for hybrid precoding problems usually lead to high computational complexity and severe system performance loss.Machine learning is adopted in the design of hybrid precoders due to its advantages of adaptive learning and decision making.Based on the fundamental theory of machine learning,a hybrid precoding algorithm using cross-entropy optimization strategy is proposed.By iteratively updating the cross-entropy loss function with robust error,the optimal hybrid precoder combination is obtained.The combination has been shown to achieve an ideal transmission sum rate and signifi cantly increase the energy effi ciency of the system.