Non-Uniform Pilot Pattern Design for Sparse Channel Estimation in MIMO-OFDM Systems

In this paper, we propose a non-uniform pilot pattern based on the cyclic difference set (CDS) for sparse channel estimation in multiple-input-multiple-output orthogonal frequency-division multiplexing (MIMO-OFDM) systems. The CDS-based pilot patterns can effectively enhance the performance of sparse channel estimation by applying compressed sensing (CS) techniques. However, the existing CDS pilot patterns are limited to single-input-single-output orthogonal frequency-division multiplexing (SISO-OFDM) systems. In MIMO-OFDM systems, as the required number of CDS increases, there will be a growing number of intersecting (overlapping) elements among different CDSs (antennas), thus introducing significant interference among different antennas. To tackle this problem, a special CDS property called identical multi-set intersection (IMI) property is disclosed which can significantly reduce the inter-antenna interference. In addition, two constructions of desired CDS collection $\bf D_{I}$ having such IMI property are presented. To further eliminate the residual interference due to the remaining intersecting elements, i.e. the exact $\lambda$ overlapping positions among all antennas, two pilot placement algorithms, i.e., zero setting (ZS) and partial orthogonal cover coding (POCC), are proposed to achieve orthogonal pilot patterns in frequency. Our simulation results show that our proposed MIMO non-uniform IMI CDS pilot patterns exhibit superior estimation accuracy, closely approaching that of SISO systems. Furthermore, our MSE simulation results also show that, under the same condition, the proposed IMI CDS pilot scheme significantly outperforms the 3GPP 5G NR MIMO-OFDM pilot schemes, offering better estimation performance while maintaining much lower generating complexity compared with other existing multi-antenna pilot schemes.