An optimal algorithm for mmWave 5G wireless networks based on neural network

Abstract

Fifth generation (5G) wireless networks are based on the use of spectrum blocks above 6 GHz in the millimeter wave (mmWave) range to increase throughput and reduce the overall level of interference in very busy frequency bands below 6 GHz. With the global deployment of the first commercial installations of 5G, the availability of multi-Gbps wireless connections in the mmWave frequency band becomes closer to reality and opens up some unique uses for 5G. Although, mmWave communication is expected to enable high-power radio links and broadband wireless intranet, its main challenges are inherent poor propagation conditions and high transmitter-receiver coordination requirement, which prevent it from realizing its full potential. When smart reflective surfaces are used in mmWave communication, channel state information becomes complex and imprecise. In this study, a hybrid intelligent reflecting surface consisting of a large number of passive components and a small number of RF circuits is proposed as a solution. Then, an improved deep neural network (DNN)-based technique is proposed to estimate the effective channel. The proposed technique provides better channel estimation performance according to the simulation results and improves the quality of service.