Low-Complexity Optimization Methods for the Evolution of Massive MIMO Technology

Massive multiple-input multiple-output (MIMO) technology has played critical roles in fifth generation (5G) systems and is expected to form the backbone for beyond 5G systems. In this thesis, we consider low-complexity optimization methods for the evolution of massive MIMO. In particular, the studies carried out in this thesis focus on resource allocation problems in cell-free massive MIMO, multigroup multicast systems, wireless federated learning, and intelligent reflecting surfaces (IRSs) aided wireless networks. First, we present a first-order accelerated projected gradient (AccPG) method for the utility maximization, including spectral efficiency and max-min fairness, in the downlink and uplink of cell-free massive MIMO. For the uplink cell-free massive MIMO, we propose a low-complexity method based on mirror prox (MP) to further improve the convergence compared to the AccPG. For multigroup multicast cell-free massive MIMO systems, we derive a closed-form achievable rate expression, incorporating power control coefficients and solve the resulting problem using a low-complexity AccPG method. We then extend the scope of our research to the wireless federated learning (FL) systems and introduce a futuristic scenario where FL and non-FL users are jointly served. In this context, the thesis aims to address how massive MIMO can support both types of communication services. In particular, we consider a fundamental problem of maximizing the effective rate of non-FL users subject to a time constraint on FL users. To solve this problem, we apply the successive convex approximation (SCA) method where novel convex approximations are proposed. Finally, we consider emerging communication systems where data transmission is assisted by intelligent reflecting surfaces (IRSs). To great extent, IRSs are considered as another form of massive MIMO evolution, since many passive reflecting elements in IRSs can create narrow passive beams. In this new massive MIMO technology, we consider the sum rate maximization problem for multigroup multicasting. Here, we make use of an alternating projected gradient (AltPG) method to achieve convergence. Throughout the thesis, simulation results demonstrate that the proposed methods arising from our reach require much less run time than known methods while providing similar or better performances. Due to their low complexity, these methods can scale favourably with large-scale wireless systems, which has been an inherent issue of the existing studies. Thus, the contributions summed up in this thesis will have a huge impact on future research in the related field, especially towards practical implementations of massive MIMO technology.