Pilot-Aided Estimation of Millimeter Wave Sparse Massive MIMO Channels using Neural Networks

Abstract

Massive multiple-input multiple-output (MIMO) has a major role in meeting the diverse demands of high throughput, low latency and massive connectivity in current 5^th generation (5G) of wireless communication networks. Also, the utilization of unused millimeter-wave (mmWave) spectrum is another substantial prospective in resolving the spectrum congestion caused by the drastic increase in the number of network users. Integration of massive MIMO with millimeter-wave spectrum is a potential paradigm, where a large number of antenna elements in the array can be supported in small physical dimensions due to the small wavelength. However, to reap the benefits, an accurate estimate of the both up- and downlink channel at the base station (BS) is of vital importance in devising efficient signal processing techniques. In this regard, both the orthogonal pilots (OP) and superimposed pilots (SP) arrangement can be used for obtaining the channel estimates (CE). However, the SP arrangement can potentially avoid the loss in spectral efficiency incurred by the sharing of pilots among the users in OP arrangement. Moreover, the mm-wave radio propagation channels are usually encountered as sparse structured in angular domain due to dominant phenomenon of specular reflection (instead of scattering encountered in micro-wave range) and high pathloss. The proposed research work aims at devising a channel estimation method for such angular domain sparse structured mmWave massive MIMO channels aided with SPs.