Bayesian learning aided doubly-selective simultaneous sparse CSI estimation in multi-user MIMO systems relying on orthogonal time frequency space modulation

Abstract

New delay and Doppler (DD)-domain channel state information (CSI) estimation schemes are conceived for multi-user (MU)-orthogonal time frequency space (OTFS) multiple-input multiple-output (MIMO) systems, for both uplink (UL) and downlink (DL)-scenarios. An important feature of the proposed schemes is that the pilots are modulated and transmitted directly in the time-domain, leading to a low pilot-overhead, flexible training duration, and reduced complexity. These techniques, therefore, achieve a high bandwidth efficiency, since they eliminate the need for DD-domain guard bands corresponding to all the transmit antennas. Moreover, the DD-domain CSI of the MU-UL system is shown to be row-sparse in nature, while that of the MU-DL scenario exhibits block-sparsity. The multiple measurement vector (MMV)-based Bayesian learning (BL) technique for UL MIMO OTFS (MB-ULMO) is then proposed for exploiting the row-sparsity and for enhancing the UL channel estimation performance. Similarly, for the downlink, the block BL (BB) for DL MIMO OTFS (BB-DLMO) algorithm is proposed in order to exploit the block-sparsity for CSI recovery. Furthermore, the BCRLBs are derived for characterizing the MSEs of the proposed DD-domain channel estimation schemes. A low-complexity linear MMSE detector (LCLMD) is derived for data detection that is ideally suited for practical implementation since it does not require any prior information about the number of active multipath components at the receiver. Our simulation results demonstrate the improved performance of the proposed BL-based schemes both in comparison to extensions of other sparse schemes and to conventional estimation techniques.

More information

Identifiers

ORCID for Lajos Hanzo: orcid.org/0000-0002-2636-5214

Catalogue record

Export record

Altmetrics