A Compressive Sensing Based Channel Prediction Scheme with Uneven Pilot Design in Mobile Massive MIMO Systems

Massive Multiple Input Multiple Output (MIMO) is a widely used technique that can provide tremendous gain in spectral efficiency. However, the degradation of beamforming performance due to outdated Channel State Information at the Transmitter side (CSIT) induced by mobility of users, both in Time Division Duplex (TDD) and Frequency Division Duplex (FDD) modes, has been a severe problem awaiting to be solved. Many channel prediction schemes have been proposed to address this problem. Most of them focus on the prediction of future CSI on pilot symbols, then reconstruct the non-pilot symbols through conventional interpolation methods. However, the CSI cannot be simply reconstructed through interpolation in high mobility scenario because of the limitation of even pilot design - the phenomenon is named as 'Doppler aliasing' in this paper. To address this, we propose a novel uneven pilot pattern that can provide more Doppler spectrum resolution compared with even pilot which is currently used in most of communication systems. Based on the novel pilot setting, we then propose a channel prediction scheme based on the compressive sensing technique. Numerical results show our scheme can outperform the state-of-the-art algorithms, including deep neural networks and autoregression models, for about 15 percents in terms of average throughput in simulated channel generated by the COST2100 channel model.