1 January 2016

Blind distributed beamforming via matrix completion

Tsinos, C. G., Vlachos, E., Berberidis, K.
2016 IEEE 17th International Workshop on Signal Processing Advances in Wireless Communications (SPAWC)
Download PDF IEEE Xplore

Abstract

In this paper, we consider the problem of distributed beamforming for maximization of the receiver signal-to-noise­ ratio (SNR) subject to a total transmit power constraint. We investigate the case where the optimal beamforming weights are expressed based on the second-order statistics of the in­ volved channels, while the communication among the relays is interference-limited. In this context, we propose a relay­ cooperative scheme for interference minimization, where only a limited number of correlation quantities are sent to the fusion center (FC). We propose a technique which overcomes the problem of the incomplete covariance matrices via matrix completion. Through simulation results, we show that, after a number of iterations, the proposed technique converges to the true covariance matrices and thus the optimal beamformer may be computed.

Type 1
Publication 2016 IEEE 17th International Workshop on Signal Processing Advances in Wireless Communications (SPAWC)
DOI 10.1109/spawc.2016.7536750
Date January 2016
Research 6G UAVs

Key Contributions

  • Addresses the challenge of distributed beamforming with incomplete channel statistics by leveraging matrix completion techniques.
  • Proposes a low-complexity iterative method for estimating correlation matrices at the fusion center using partial measurements.
  • Demonstrates that the technique converges to the optimal beamforming solution with a limited number of relays providing direct communication.

Results & Insights

Convergence curves for matrix completion of correlation matrices R(t) and G(t) versus time, showing NMSE reduction for different numbers of known entries.
Convergence curves for matrix completion of correlation matrices R(t) and G(t) versus time, showing NMSE reduction for different numbers of known entries.

These results demonstrate that the proposed matrix completion technique achieves near-optimal performance with significantly fewer known entries (34% for K=10), highlighting its efficiency in recovering the full correlation matrices needed for optimal beamforming.

Separate NMSE convergence of correlation matrices R(t) (top) and G(t) (bottom), comparing full-entry recovery against logarithmic subsampling strategies.
Separate NMSE convergence of correlation matrices R(t) (top) and G(t) (bottom), comparing full-entry recovery against logarithmic subsampling strategies.

The results show that both matrices converge reliably using only O(K log K) known entries, validating the scalability and practicality of the matrix completion approach for distributed beamforming.

Research 6G UAVs