Supervisors info:
Άρης Μουστάκας, Αναπληρωτής Καθηγητής, Τμήμα Φυσικής, ΕΚΠΑ (κύριος επιβλέπων)
Άννα Τζανακάκη, Επίκουρη Καθηγήτρια, Τμήμα Φυσικής, ΕΚΠΑ (μέλος επιτροπής)
Ανδρέας Πολύδωρος, Καθηγητής, Τμήμα Φυσικής, ΕΚΠΑ (μέλος επιτροπής)
Summary:
In the present thesis what is investigated is a new network architecture which by abolishing the conventional cellular system (Cell-Free) promises not only to provide full satisfaction of the standards set by the current 5G networks, but also to create the appropriate conditions for the reception of the next generations of networks. However, as any modern wireless communication system, Cell-Free requires the utilization of the possibilities provided by the technologies of multiple antennas, particularly that of the Massive MIMO architecture. Hence, for a better understanding of Cell-Free Massive MIMO reasoning, after an introduction to the most important problems that influence each wireless propagation channel, an extensive presentation of the potential entailed in each co-located MIMO topology is provided.
In particular, Chapter 2 referes to the concepts of spatial diversity, spatial multiplexing and array gain, properties that make each MIMO layout a powerful compensating factor against the problems of reduced channel capacity and of general link instability. The essential role of MIMO topology is highlighted through the relevant modeling of the system, which functions as the cornerstone for the subsequent construction of more complex arrangements. Additionally, the simple point-to-point SU-MIMO is adapted to mobile communications through the introduction of MU-MIMO, which embodies the properties of the clasic MIMO in practice. This chapter concludes with a reference to the methods of retrieving and using the CSI matrix, as well ass the beamforming technique, which is a vital element for any antenna array.
Chapter 3 referes to the specifications and demands of 5G networks through the quantification of certain indicators, which are related to both the densification of cells and the increase of spectral efficiency (SE). In the effort to increase the second indicator, the Massive MIMO topology is fully engaged thus achieving spectacular results in terms of quality of service (QoS). The augmentation of the cell density is accomplished by the architecture of the heterogeneous networks (HetNets), which alongside with the Massive MIMO topology, constitutes even today the tip of the technological spear. In his way, the stage is set for future comparison between Cell-Free Massive MIMO and the HetNet layouts. Last but not least, an energy analysis associated with the above mentioned topologies is included, adding to this sector a practical benchmark for subsequent simulations.
In Chapter 4, after a brief reference to the problems faced by current cellular (co-located) Massive MIMO arrangements, the Cell-Free Massive MIMO topologies are introduced, mainly focusing on the features that render them fully competitive and innovative. In particular, after an extensive modeling and deriviation of closed-form rate expresions for both the uplink and the downlink schemes, a practical comparison of the Cell-Free Massive MIMO architecture with the optimal one (until today) of small cells is following, considering only the case of single-antenna access points (APs) and user-equipments (UEs). For the completeness of the analysis, various scenarios are considered. These scenarios include the existence of imperfect CSI matrix, of non-orthogonal pilot signals (pilot contamination), of the max-min power control strategy and of two different pilot assignment algorithms. The chapter concludes with the modeling and the simulation of Cell-Free Massive MIMO systems in a context with multi-antenna APs and multi-antenna UEs examining the enhancing or degrading nature of this addition.
Following, Chapter 5 presents several strategies for improving the performance of Cell-Free Massive MIMO systems. Among these, the 4 levels of cooperation between APs and CPU during the uplink process are described, considering at the same time an optimal combining technique. This aims to fully understand the correct design methodology of non-cellular Massive MIMO layouts, so that they are able to provide higher levels of performance compared to co-located Massive MIMO topologies in every scenario. Regarding the downlink process, a number of improving and innovative precoding algorithms are proposed, which, acting suppressively for the generated interferences, provide quite satisfactory rates to the system. The chapter ends with the adaptation of the NOMA strategy to the date of Cell-Free Massive MIMO layouts, enhancing the performance of the system even in high date traffic cases.
Chapter 6 addresses the issue of Cell-Free Massive MIMO fronthaul signaling, that is the complexity that develops in it due to the transfer of large volumes of data. The solution proposed includes the serial connection of the fronthaul part of the layout through the innovative topology of radio stripes, combined with a serial proccessing algorithm, which performes a gradual reduction of errors thus making it entirely competitive. The chapter continues with a complete modeling of the Cell-Free Massive MIMO system, this time taking into account the existence of limited fronthaul capacity and of hardware impairments, aiming at the most faithfull identification of its implementation conditions within the imperfections of reality.
The last chapter features an energy profile, as it includes the modeling of the Cell-Free Massive MIMO system with the sole purpose of its studying in the forementioned field. A key role is played in the first paragraph where the energy efficiency (EE) of cellular and Cell-Free Massive MIMO layouts are compared after taking into account three different implementation scenarios, as well as two standard power control algorithms. For the full depiction of the possibilities of the proposed topology, a more dedicated analysis takes place, which includes the existence of the pilot contamination phenomenon and two basic AP selection schemes, resulting in a non-convex energy minimization problem. Lastly, this thesis is completed by providing the results of energy content simulations of Cell-Free Massive MIMO topologies in which channel correlation is observed alongside with limited fronthaul capacity in combination with hardware impairments.
