Fast Algorithms for Joint Multicast Beamforming and Antenna Selection in Massive MIMO

Abstract

Massive MIMO is currently a leading physical layer technology candidate that can dramatically enhance throughput in 5G systems, for both unicast and multicast transmission modalities. As antenna elements are becoming smaller and cheaper in the mmW range compared to radio frequency (RF) chains, it is crucial to perform antenna selection at the transmitter, such that the available RF chains are switched to an appropriate subset of antennas. This paper considers the joint problem of multicast beamforming and antenna selection for a single multicast group in massive MIMO systems. The prior state-of-art for this problem relies on semi-definite relaxation (SDR), which cannot scale up to the massive MIMO regime. A successive convex approximation (SCA) based approach is proposed to tackle max-min fair joint multicast beamforming and antenna selection. The key idea of SCA is to successively approximate the non-convex problem by a class of non-smooth, convex optimization problems. Two fast and memory efficient first-order methods are proposed to solve each SCA subproblem. Simulations demonstrate that the proposed algorithms outperform the existing state-of-art approach in terms of solution quality and run time, in both traditional and especially in massive MIMO settings.

First-Order Fast Algorithm for Structurally Optimal Multi-Group Multicast Beamforming in Large-Scale Systems

“…The main complexity in the SCAbased algorithms is from the interior-point method (IPM) used to solve the convex subproblem at each SCA iteration. For the singlegroup case, [19,20] have developed first-order algorithms to solve the convex subproblems, with much lower complexity than the IPM method. Nevertheless, the common issue of these methods is that the complexity grows in polynomial time with the number of antennas, making them still computationally heavy for large-scale massive MIMO systems.…”

“…However, in numerical computation, the IPM-based SCA method is still adopted, which may not be efficient as the number of users becomes large. Also, the first-order method in [19,20] for the single-group case is not applicable to the multigroup case. Thus, our goal is to develop a fast algorithm based on the optimal solution structure for large-scale systems.…”

User Oriented Transmit Antenna Selection in Massive Multi-User MIMO SDR Systems

“…In [ 16 ], the authors proposed a TxAS algorithm to maximize the system energy efficiency, and a decent approximation of the distribution of the mutual information was obtained for the AS aided MIMOs. Research in [ 17 ] proposed two successive convex approximation (SCA) based algorithms to address the two different forms of each non-smooth, convex subproblem that showed superior performance to TxAS aided massive MIMO systems with multicast beamforming. Additionally, under the condition of limited backhaul capacity in distributed MIMO systems, joint TxAS, and user scheduling algorithms were proposed in [ 18 ].…”

Optimal Solutions for Joint Beamforming and Antenna Selection: From Branch and Bound to Graph Neural Imitation Learning

“…The vast majority of the literature tackles this problem using continuous programming-based approximations. For example, [5]- [7], [10] used convex and nonconvex group sparsity-promoting regularization to encourage turning off antenna elements. However, the continuous approximations are often NP-hard problems as well (especially when the sparsity promotion is done via nonconvex quasi-norms as in [5]), and thus it is unclear if they can solve the problem of interest optimally.…”

“…Our design leverages problem structures of unicast BF and RBF, which allows for branching only on a subset of the optimization variables-thereby having reduced complexity and being effective. Unlike continuous optimization-based approximations in [5]- [7], [10] whose solutions are often suboptimal or infeasible, the proposed B&B is guaranteed to return an optimal solution. • An ML-based Acceleration Scheme.…”