Beam selection for performance-complexity optimization in high-dimensional MIMO systems

Abstract

Millimeter-wave (mm-wave) communications sys tems offer a promising solution to meeting the increasing data demands on wireless networks. Not only do mm-wave systems allow orders of magnitude larger bandwidths, they also create a high-dimensional spatial signal space due to the small wavelengths, which can be exploited for beamforming and multiplexing gains. However, the complexity of digitally processing the entire high-dimensional signal is prohibitive. By exploiting the inherent channel sparsity in beamspace due to highly directional propagation at mm-wave, it is possible to design near-optimal transceivers with dramatically lower complexity. In such beamspace MIMO systems, it is first necessary to determine the set of beams which define the low-dimensional communication subspace. In this paper, we address this beam selection problem and introduce a simple power-based classifier for determining the beamspace sparsity pattern that characterizes the communication subspace. We first introduce a physical model for a small cell which will serve as the setting for our analysis. We then develop a classifier for the physical model, and show its optimality for a class of ideal signals. Finally, we present illustrative numerical results and show the feasibility of the classifier in mobile settings.

Millimeter Wave Communications for Future Mobile Networks

“…For users with low inter-beam interference, it directly selects the beams with large power, while for users with severe inter-beam interference, a low-complexity incremental algorithm was proposed. More beam-selection schemes for N3 can be found in [170], [171].…”

Low‐complexity heuristics to beam selection and rate adaptation in sparse massive MIMO systems

“…Following some previous works, the extension for a rectangular uniform planar array using the same DFT matrix for all UEs can be done. However, a Kronecker product is used to represent the azimuth and elevation spatial frequencies in two‐dimensional arrays . Furthermore, the array structure does not affect the precoder design (based on beam selection).…”

“…Following the works of Hogan and Sayeed and Rappaport et al, in some scenarios (eg, in millimeter waves), the multipaths are expected to form a sparse set of single‐bounce components. In this context, the poor scattering effects can result in the virtual angular domain with sparse channel matrix representation .…”

“…This assumption reduces the complexity of the model, and it is known that the channel for each UE can be represented using the same basis when M T is sufficiently large. 8,15,26,27 Following some previous works, 16,28 the extension for a rectangular uniform planar array using the same DFT matrix for all UEs can be done. However, a Kronecker product is used to represent the azimuth and elevation spatial frequencies in two-dimensional arrays.…”

“…Recent studies have demonstrated that, as the spatial dimension increases, the physical MIMO channels exhibit poor scattering. 9,[15][16][17] This is the case in macrocell urban environments, where the propagation links between the BS and UE are often blocked by large buildings or when clusters of multipaths are shared by the same MUs. 7,[18][19][20] In this paper, we propose low-complexity heuristics to beam selection and rate adaptation in sparse massive MIMO systems.…”