Machine Learning-based Adaptive Connection Hybrid Precoding for mmWave Massive MIMO Systems

Millimeter-wave massive MIMO system hybrid precoding is one of the key technologies to increase the capacity of wireless communication systems and reduce the number of RF chains used, but still requires a large number of high-precision phase shifters to achieve array gain. To solve this problem, in this paper, first, by maximizing the received signal power of each user, the matching relationship between the RF chain and the base station antenna in the adaptive connection structure with one-bit quantized phase shifter is obtained, and then the adaptive cross entropy optimization method based on machine learning is innovatively applied to adaptive connection structure in hybrid precoding. By reducing the cross-entropy and adding constant smoothing parameter to ensure convergence, the probability distribution is adaptively updated to obtain an almost optimal hybrid precoder. Finally, simulations verify the feasibility and satisfactory achievable sum-rate of the proposed scheme. It has better achievable sum-rate performance compared with other hybrid precoding schemes with the same hardware complexity.