Loading [a11y]/accessibility-menu.js
Attention Neural Network for Downlink Cell-Free Massive MIMO Power Control | IEEE Conference Publication | IEEE Xplore

Attention Neural Network for Downlink Cell-Free Massive MIMO Power Control

Publisher: IEEE

Abstract:

The downlink power control is challenging in a cell-free massive multiple-input multiple-output (CFmMIMO) system because of the non-convexity of the problem. This paper p...View more

Abstract:

The downlink power control is challenging in a cell-free massive multiple-input multiple-output (CFmMIMO) system because of the non-convexity of the problem. This paper proposes a computationally efficient deep-learning algorithm to solve the max-min power control optimization problem subject to power constraints. To solve this problem, it presents an attention neural network(ANN) composed using the masked multi-head attention network modules, which are building blocks of the popular transformer neural network. The ANN solves the downlink power control problem of CFmMIMO in the presence of pilot contamination (non-orthogonal pilot sequences). The paper first translates the constrained optimization problem to an unconstrained one parameterized by the weights of the ANN. These weights are trained in an unsupervised fashion. The performance of the ANN power control algorithm is demonstrated using numerical simulations. The paper also provides a computational complexity analysis of the method.
Date of Conference: 31 October 2022 - 02 November 2022
Date Added to IEEE Xplore: 07 March 2023
ISBN Information:

ISSN Information:

Publisher: IEEE
Conference Location: Pacific Grove, CA, USA

I. Introduction

Cell-free massive multiple-input multiple-output (CFm-MIMO) performs multiuser multiple-input multiple-output (MIMO) communications across distributed antennas or access points (APs). Decentralizing the antennas improves channel orthogonality and provides higher rates than colocated antennas, thanks to the differences in path loss between the users [1]–[7]. The time-division duplex (TDD) based CFmMIMO system with a large number of distributed antennas cooperating to serve a fewer number of users has been proposed in [8]. The CFmMIMO provides uniformly good service throughout the area of coverage. However, an efficient uplink/downlink power control algorithm is essential for achieving uniformly good service to all the users. Power control in CFmMIMO is challenging in general because of the computational complexity involved in solving the non-convex max-min fairness maximization problem [8], [9]. Specifically, designing a computationally efficient power control algorithm for the downlink signal at the central processing unit (CPU) is very challenging for CFmMIMO because of the number of optimization parameters involved in the downlink power control [9].

References

References is not available for this document.