Abstract

Along with its primary mission in fulfilling the communication needs of humans as well as intelligent machines, fifth generation (5G) and beyond networks will be a virtual fundamental component for all parts of life, society, and industries. These networks will pave the way towards realizing the individuals’ technological aspirations including holographic telepresence, e-Health, pervasive connectivity in smart environments, massive robotics, three-dimensional unmanned mobility, augmented reality, virtual reality, and internet of everything. This new era of applications brings unprecedented challenging demands to wireless network, such as high spectral efficiency, low-latency, high-reliable communication, and high energy efficiency. One of the major technological breakthroughs that has recently drawn the attention of researchers from academia and industry to cope with these unprecedented demands of wireless networks is the cell-free (CF) massive multiple-input multiple-output (mMIMO) systems. In CF mMIMO, a large number of spatially distributed access points are connected to a central processing unit (CPU). The CPU operates all APs as a single mMIMO network with no cell boundaries to serve a smaller number of users coherently on the same time-frequency resources. The system has shown substantial gains in improving the network performance from different perspectives, especially for cell-edge users, compared it other candidate technologies for 5G networks, \ie co-located mMIMO and small-cell (SC) systems. Nevertheless, the full picture of a practical scalable deployment of the system is not clear yet. In this thesis, we provide more in-depth investigations on the CF mMIMO performance under various practical system considerations. Also, we provide promising solutions to fully realize the potential of CF mMIMO in practical scenarios. In this regard, we focus on three vital practical challenges, namely hardware and channel impairments, malicious attacks, and limited-capacity fronthaul network.

Regarding the hardware and channel impairments, we analyze the CF mMIMO performance under such practical considerations and compare its performance with SC systems. In doing so, we consider that both APs and user equipment (UE)s are equipped with non-ideal hardware components. Also, we consider the Doppler shift effect as a source of channel impairments in dynamic environments with moving users. Then, we derive novel closed-form expressions for the downlink (DL) spectral efficiency of both systems under hardware distortions and Doppler shift effect. We reveal that the effect of non-ideal UEs is more prominent than the non-ideal APs. Also, while increasing the number of deployed non-ideal APs can limit the hardware distortion effect in CF mMIMO systems, this leads to an extra performance loss in SC systems. Besides, we show that the Doppler shift effect is more harsh in SC systems. In addition, the SC system operation is more suitable for low-velocity users, however, it is more beneficial to adopt CF mMIMO system for network operation under high-mobility conditions. Capitalizing on the latter, we propose a hybrid CF mMIMO/SC system that can significantly outperforms both CF mMIMO and SC systems by providing different mobility conditions with high data rates simultaneously.

Towards a further improvement in the CF mMIMO performance under high mobility scenarios, we propose a novel framework to limit the performance degradation due to the Doppler shift effect. To this end, we derive novel expressions for tight lower bound of the average DL and uplink (UL) data rates. Capitalizing on the derived analytical results, we provide an analytical framework that optimizes the frame length to minimize the Doppler shift effect on DL and UL data rates according to some criterion. Our results reveal that the optimal frame lengths for maximizing the DL and UL data rates are different and depend mainly on the users' velocities. Besides, adapting the frame length according to the velocity conditions significantly limits the Doppler shift effect, compared to applying a fixed frame length.

To empower the CF mMIMO systems with secure transmission against malicious attacks, we propose two different approaches that significantly increases the achievable secrecy rates. In the first approach, we introduce a novel secure DL transmission technique that efficiently limits the eavesdropper (Eve) capability in decoding the transmitted signals to legitimate users. Differently, in the second approach, we adopt the distinctive features of Reconfigurable intelligent surfaces (RIS)s to limit the information leakage towards the Eve.

Regarding the impact of limited capacity of wired-based fronthaul links, we drive the achievable DL data rates assuming two different CF mMIMO system operations, namely, distributed and centralized system operations. APs and CPU are the responsible entities for carrying out the signal processing functionalities in the distributed and centralized system operations, respectively. We show that the impact of limited capacity fronthaul links is more prominent on the centralized system operation. In addition, while the distributed system operation is more preferable under low capacities of fronthaul links, the centralized counterpart attains superior performance at high capacities of fronthaul links. Furthermore, considering the distributed and centralized system operations, and towards a practical and scalable operation of CF mMIMO systems, we propose a wireless-based fronthaul network for CF mMIMO systems under three different operations, namely, microwave-based, mmWave-based, and hybrid mmWave/microwave. Our results show that the integration between the centralized operation and the hybrid-based fronthaul network provides the highest DL data rates when APs are empowered with signal decoding capabilities. However, integrating the distributed operation with the microwave-based fronthaul network achieves ultimate performance when APs are not supported with decoding capabilities.