PhD defence of Seyyed Saleh Hosseini - Studies in Cell-Free Massive-MIMO: Green Power Allocation, Physical Security, and DoA Estimation

Abstract

Cell-free massive MIMO (m-MIMO) has been designated as a key enabling technology for beyond the fifth generation (B5G) and the sixth generation (6G) of wireless communication networks. It is essentially a distributed heterogeneous network in which a large number of access points (APs), scattered over a geographical area and coordinated by a central processing unit (CPU), serve multiple users without being bounded by cells. Cell-free m-MIMO, as a state-of-the-art technology, offers several advantages, including: high spectral efficiency (SE), large-scale diversity gain, and avoiding the need for frequent handovers.

Nevertheless, the practical deployment of cell-free m-MIMO poses several signal processing challenges at the physical layer. For instance, the total power consumption for the communication links can reach such a high level as to defeat the SE gains. As a distributed network, cell-free m-MIMO suffers potentially from physical security issues, especially the pilot spoofing attack, which is launched by an eavesdropper (Eve) to overhear an intended user. Finally, the quality of certain channel parameters, required for signal processing tasks at the CPU can be severely affected due to the use of low-resolution quantizers on the backhaul links between the APs and the CPU.

In this thesis, we respectively address and propose novel solutions to the above issues in three parts. Specifically, we commence with the problem of downlink power allocation in a cell-free m-MIMO system under SE constraints for the users. The power allocation is formulated as an optimization problem where the aim is to maximize the sum SE as the objective function, while limiting the transmission power of APs and imposing lower and upper bounds on the achievable SEs of different users. While the problem is non-convex, an efficient solution approach is developed through the use of bounding and relaxation techniques. Interestingly, it reveals that each user is allocated a fraction of available power proportional to its required data rate, which in turn, leads to a significant reduction in total power consumption. Besides, the quality of service can be enhanced for users who require high SE but are located at the periphery of the network coverage area.

In the second part of the thesis, we propose two novel methods based on the log-likelihood ratio test (LLRT), one in a centralized and the other in a decentralized fashion, to cope with the problem of pilot spoofing attack in a cell-free m-MIMO system. The methods take advantage of a special protocol in which the legitimate users switch to an off-mode irregularly, without significantly affecting the spectral efficiency of the data transmission. The protocol is applicable to environments with low to moderate mobility but can be extended to high mobility through a simple rearrangement of available pilot sequences among users. The detection performances of the proposed methods are mathematically analyzed and their validity is confirmed via simulations. Moreover, the proposed methods significantly outperform benchmark approaches from the recent literature in terms of detection and false alarm probabilities, while requiring low fronthaul overhead.

In the last part of the thesis, we consider the problem direction of arrival (DoA) estimation in cell-free m-MIMO and propose a new method based on cutting-edge deep neural network (DNN) technology. To train the DNN, a special feature set is proposed as obtained from the first superdiagonal entries of the spatial correlation matrix. This selection of features makes it possible to employ a DNN with only a few low-dimensional layers, which considerably speeds up training and processing. The proposed method offers a high resolution and a significant reduction in terms of processing time compared to the established approaches in the literature.