Beam Selection in mm-Wave Multiuser MIMO Systems Using Compressive Sensing

Abstract

In this paper, we study beam selection for millimeterwave (mm-wave) multiuser multiple input multiple output (MIMO) systems where a base station (BS) and users are equipped with antenna arrays. Exploiting a certain sparsity of mm-wave channels, a low-complexity beam selection method for beamforming by low-cost analog beamformers is derived. It is shown that beam selection can be carried out without explicit channel estimation using the notion of compressive sensing (CS). Due to various reasons (e.g., the background noise and interference), some users may choose the same BS beam, which results in high inter-user interference. To overcome this problem, we further consider BS beam selection by users. Through simulations, we show that the performance gap between the proposed approach and the optimal beamforming approach, which requires full channel state information (CSI), becomes narrower for a larger number of users at a moderate/low signal-to-noise ratio (SNR). Since the optimal beamforming approach is difficult to be used due to prohibitively high computational complexity for large antenna arrays with a large number of users, the proposed approach becomes attractive for BSs and users in mm-wave systems where large antenna arrays can be employed.

Hybrid Beamforming in mm-Wave MIMO Systems Having a Finite Input Alphabet

“…We then obtain the constrained solution via matrix decomposition [11] in order to obtain an equal gain element matrix and a unit norm matrix, which are used as analog and digital precoding/combining matrices, respectively. Furthermore, motivated by the recent developments in directional beamforming (DBF) in the context of mm-wave communication [17]- [19], we propose a low-complexity gradient-ascent aided DBF (GA-DBF) that strikes a beneficial trade-off between the complexity imposed and the performance attained. 2) Ideally, assuming perfect channel state information (CSI) at the receiver, the precoding/combining matrices are computed as stated above under 1) and then both the associated analog as well as digital precoding matrices are conveyed to the transmitter.…”

“…Motivated by the simplicity of directional beamforming solutions [17]- [19], we consider a DFT-based directional codebook for analog BF matrices and for a gradient-ascent based iterative solution conceived for digital BF matrices. The proposed low-complexity solution is presented in Algorithm 2.…”

“…When the number of channel paths increases, the number of partition cells has to be exponentially increased in order have a reasonable quantization resolution, which in turn makes the VQ codebook design computationally complex. Motivated by the low-complexity directional beamforming solutions of [17]- [19], we propose a VQ codebook based solution by partitioning the set of baseband channels, which are obtained after employing DFT codebook based analog BF, as seen in Algorithm 2. This scheme is referred to as the VQ codebook aided directional beamformer (VQ-DBF).…”

Iterative Matrix Decomposition Aided Block Diagonalization for mm-Wave Multiuser MIMO Systems

“…In [17], the authors propose both single and multi-user based analog beam selection, while considering both minimum mean square error (MMSE) and ZF based solutions for digital precoding. In [18], the authors propose beam selection algorithms based on the compressive sensing approach, which rely on the orthogonal matching pursuit (OMP) algorithm [23] for finding dominant beams. In [19], the authors propose a two-stage algorithm in which the first stage obtains the analog BF matrices and the second stage obtains the digital BF matrix based on the classic zero-forcing approach.…”

“…Note that this approach is fundamentally different from the existing BD schemes [24], [25], where the precoders are designed only to block-diagonalize the multi-user channel without any explicit motivation to enhance the signal power. The BD schemes in the existing mm-wave communication literature are employed in the digital domain, while the RF precoding is generally employed using beam-steering [18], [19]. Explicitly, the BF and BD are performed independently in the analog and digital domains, respectively.…”