Channel Estimation And Pilots Optimization For Massive MIMO-OFDM Systems Based On Structured Compressive Sensing

Large-scale MIMO-OFDM system has become one of the key technologies in 5G for its higher channel capacity and higher energy efficiency.This paper addresses the sparse channel estimation and pilot optimization problems in large-scale MIMO-OFDM system from the perspective of structured compressive sensing.The main contributions are as follows:(1)In large-scale MIMO-OFDM system,if the pilots in different transmit antennas are superimposed,the problem of sparse channel estimation can be modeled as the reconstruction problem of structured compressive sensing.Thereby,the relation between the sparse channel estimation in large-scale MIMO-OFDM system and the structured compressive sensing is established.(2)Considering the pilot design will involve two key factors,i.e.the pilot locations and pilot symbols.To improve the performance of structured compressive sensing-based channel estimation,a mutual coherence-related criterion is proposed to optimize the pilot locations firstly.And a random research algorithm is provided to obtain optimized pilot locations.Simulation results show that,as compared with other randomly generated pilot locations,employing the pilot locations obtained by proposed optimizing algorithm can reduce the error of channel estimation obviously,and thus improve the channel estimation performance.The performance gain obtained by employing optimized pilot locations is about 2~4dB.(3)Considering the pilot locations and pilot symbols simultaneously,the mutual coherence-related criterion is proposed to optimize pilots for the pilot locations and pilot symbols.Also a random research algorithm is provided to obtain optimized pilots.Simulation results show that,as compared with other randomly generated pilots,employing the pilots obtained by proposed optimizing algorithm can reduce the error of channel estimation obviously,and thus improve the channel estimation performance.The performance gain obtained by employing optimizing pilots is about 2~5dB.Simulation results also show that,the performance of optimizing pilot locations and pilot symbols simultaneously is superior to that of optimizing only pilot locations in terms of the MSE of channel estimation.