Abstract
We propose a new antenna selection scheme for a massive MIMO system with a single user terminal and a base station with a large number of antennas. We consider a practical scenario where there is a realistic correlation among the antennas and imperfect channel estimation at the receiver side. The proposed scheme exploits the sparsity of the channel matrix for the effective selection of a limited number of antennas. To this end, we compute a sparse channel matrix by minimising the mean squared error. This optimisation problem is then solved by the well-known orthogonal matching pursuit algorithm. Widely used models for spatial correlation among the antennas and channel estimation errors are considered in this work. Simulation results demonstrate that when the impacts of spatial correlation and imperfect channel estimation introduced, the proposed scheme in the paper can significantly reduce complexity of the receiver, without degrading the system performance compared to the maximum ratio combining.
1. Introduction
Massive MIMO (Multiple-Input Multiple-Output) is a cornerstone of 5G and beyond, promising massive throughput by packing dozens or hundreds of antennas at a base station. However, hardware cost and complexity, along with physical constraints like spatial correlation and imperfect Channel State Information (CSI), present significant implementation challenges. This paper addresses the problem of selecting a subset of antennas to reduce hardware costs (RF chains) without sacrificing performance in realistic, non-ideal channel conditions.
2. Methodology
The proposed scheme frames antenna selection as a sparse approximation problem. By defining the Mean Squared Error (MSE) between the transmitted and reconstructed signals, the authors introduce a sparse selection vector where zero entries represent unselected antennas. The selection problem is solved using the Orthogonal Matching Pursuit (OMP) algorithm, an iterative greedy approach that identifies the most significant antennas one by one. The study incorporates the Kronecker model for spatial correlation and an MMSE-based model for channel estimation errors, ensuring the mathematical validity of the approach through Cholesky decomposition.
3. Results and Analysis
The performance of the OMP-based selection scheme was evaluated through Monte Carlo simulations for base stations with $M = 16, 64,$ and $128$ antennas.
As illustrated in Figure 1, the proposed scheme maintains performance levels very close to the full Maximum Ratio Combining (MRC) baseline across a range of Signal-to-Noise Ratios (SNR), even under high correlation and estimation error. A critical finding is the relationship between the number of selected antennas ($K_s$) and the resulting BER.
Figure 2 demonstrates that for a 64-antenna system, selecting just 32 antennas ($K_s=32$) yields a BER nearly indistinguishable from the full MRC scheme. This "half-power sweet spot" indicates that redundancy in correlated channels can be effectively exploited to reduce hardware complexity ($O(K_s^2 M)$) by minimizing the number of required physical RF chains.
4. Conclusion
The proposed OMP-based antenna selection scheme offers a robust solution for massive MIMO systems, achieving near-optimal performance with significantly reduced hardware requirements. While the current work focuses on single-user uplinks, future research will explore extensions to multi-user MIMO (MU-MIMO) scenarios and faster selection methods to further reduce latency.
Summary updated