Modified CS-based downlink channel estimation with temporal correlation in FDD massive MIMO systems

Abstract

In order to fully utilize the spatial multiplexing gains and array gains of massive multiple-input multiple-output (MIMO), it is essential to obtain accurate channel state information at the transmitter (CSIT). However, conventional CSIT estimation approaches are not suitable for frequency-division duplexing (FDD) massive MIMO systems due to the prohibitively large training and feedback overhead. Recently, the compressive sensing (CS) technique is proposed to reduce the overhead for CSIT acquisition. In this paper, we consider CS-based channel estimation schemes with temporal correlation. First, we discuss the existing modified subspace pursuit (M-SP) algorithm which can exploit the prior channel support information to enhance the current CSIT estimation performance. To tolerate model mismatch, the conservative M-SP algorithm is further discussed. Moreover, we propose a new adaptive M-SP algorithm which not only exploits the temporal correlation of the massive MIMO channels, but also has the built-in learning capability to adapt to the appropriate prior channel support quality parameter. The proposed adaptive M-SP algorithm can obtain the successful CSIT recovery in the case of model mismatch. Simulation results show that the proposed algorithm has substantial performance gain over conventional algorithms and is very robust to model mismatch.