The broadcast nature of wireless medium has made information security as one of the most important and critical issues in wireless systems. Physical layer security, which is based on information-theoretic secrecy concepts, can be used to secure the wireless channels by exploiting the noisiness and imperfections of the channels. Massive multiple-input multiple-output (MIMO) systems, which are equipped with very large antenna arrays at the base stations, have a great potential to boost the physical layer security by generating the artificial noise (AN) with the exploitation of excess degrees-of-freedom available at the base stations. In this thesis, we investigate physical layer security provisions in the presence of passive/active eavesdroppers for single-hop massive MIMO, dual-hop relay-assisted massive MIMO and underlay spectrum-sharing massive MIMO systems. The performance of the proposed security provisions is investigated by deriving the achievable rates at the user nodes, the information rate leaked into the eavesdroppers, and the achievable secrecy rates. Moreover, the effects of active pilot contamination attacks, imperfect channel state information (CSI) acquisition at the base-stations, and the availability of statistical CSI at the user nodes are quantified. The secrecy rate/performance gap between two AN precoders, namely the random AN precoder and the null-space based AN precoder, is investigated. The performance of hybrid analog/digital precoding is compared with the full-dimensional digital precoding. Furthermore, the physical layer security breaches in underlay spectrum-sharing massive MIMO systems are investigated, and thereby, security provisions are designed/analyzed against active pilot contamination attacks during the channel estimation phase. A power-ratio based active pilot attack detection scheme is investigated, and thereby, the probability of detection is derived. Thereby, the vulnerability of uplink channel estimation based on the pilots transmitted by the user nodes in time division duplexing based massive MIMO systems is revealed, and the fundamental trade-offs among physical layer security provisions, implementation complexity and performance gains are discussed.

Par la révolution du Cloud Computing et des Smartphones, une quantité énorme de données devrait traverser le réseau chaque seconde où la plupart de ces données sont fournies par des mobiles utilisant des services Internet. La croissance rapide de la bande passante et des demandes de QoS rend les réseaux mobiles du 4ème G insuffisants. Le système de prochaine génération doit avoir un taux de sommation de 100Mbps à 1Gbps par terminal utilisateur (UT), avec une densité de connexion supérieure à 1M connexion / Km2, la mobilité des véhicules à grande vitesse jusqu'à 500 km / h et une fin à la fin (E2E) retardent moins de 10 ms. Un candidat prometteur qui peut répondre à ces demandes est le système sans fil à multiples sorties multiples (MIMO) Multi-Cell Multi-Cell. Cependant, la capacité Massive MIMO est délimitée par l'Inter-cell Interference (ICI) en raison de la réutilisation du pilote et, par conséquent, de la contamination du pilote. Dans cette thèse, nous étudions la contamination du pilote de liaison montante dans le système de formation à la division temporelle (TDD) des réseaux sans fil MIMO massifs. En supposant un canal de décoloration, l'intervalle de cohérence sera temporairement limité, où l'estimation du canal, la réception des symboles et le précodage des symboles doivent être effectués dans le même intervalle. Cela dit, la longueur du pilote de formation est limitée. De même, le nombre de terminaux de l'utilisateur (UT) par zone d'interférence est également limité. Inspiré par la variation de la taille de l'intervalle de cohérence parmi les UT, cette recherche présente deux nouvelles contributions indépendantes pour faire face à la contamination pilote de liaison montante dans le MIMO massif. La première contribution répertorie la région de couverture de la cellule de base (BS) dans une carte d'information d'état de chaîne (CSI). Cette carte est créée et mise à jour à l'aide d'un algorithme spécial d'apprentissage machine, et elle est exploitée pour prédire UT CSI au lieu d'estimer ses canaux. Compte tenu de cela, la formation des pilotes aériens et de liaison montante est considérablement réduite. La deuxième contribution classe les UT en fonction de la taille de leur intervalle de cohérence de canal. En outre, nous appliquons une technique de changement de pilote pour déplacer des pilotes similaires vers différentes positions temporelles (qui sont considérées comme vides en raison de trames TDD pilotes vides). Les résultats de la simulation montrent une augmentation à l'échelle de la performance du MIMO massif, en particulier dans la performance de l'efficacité énergétique et spectrale, UT par cellule et taux d'addition. En particulier, la troisième contribution évolue le MIMO massif multi-cellulaire à une performance de cellule unique et même surmonté un simple énorme conventionnel dans l'efficacité énergétique et UT par cellule. / By the revolution of Cloud Computing and Smartphones, an enormous amount of data should traverse the network every second where most of this data are delivered by mobiles using internet services. The fast growth in bandwidth and QoS demands makes the 4th G mobile networks insufficient. The next generation system must afford a sum rate from 100Mbps up to 1Gbps per User Terminal (UT), with a connection density that exceeds 1M connection/Km2, the mobility of high-speed vehicles up to 500 km/hr and an End to End (E2E) delay less than 10ms. A promising candidate that can offer those demands is the Multi-User Multi-Cell Massive Multiple-Input Multiple-Output (MIMO) wireless system. However, Massive MIMO capacity is upper bounded by the Inter-cell Interference (ICI) due to pilot reuse and thus, pilot contamination. In this thesis, we investigate the uplink pilot contamination in Time Division Duplexing (TDD) training scheme of massive MIMO wireless networks. Assuming block-fading channel, the coherence interval will lag for a limited duration, where channel estimation, symbol reception, and symbol precoding must be done within the same interval. Having said that, the training pilot length is limited. Likewise, the number of User Terminal’s (UT’s) per interference region is also limited. Inspired by the variation of coherence interval size among UT’s, this research introduces two independent novel contributions to deal with uplink pilot contamination in massive MIMO. The first contribution maps the Base Station (BS) cell coverage region into a Channel State Information (CSI) Map. This map is created and updated using a special machine-learning algorithm, and it is exploited to predict UT CSI instead of estimating their channels. In view of this, training overhead and uplink pilots are reduced significantly. The second contribution classifies UT’s based on the size of their channel coherence interval. Furthermore, we apply a pilot shifting technique to shift similar pilots to different time position (that considered empty due to empty pilot TDD frames). Simulation results show a scaled increase in the performance of massive MIMO especially in the performance of energy and spectral efficiency, UT per cell and sum-rate. In particular, the third contribution evolves multi-cell massive MIMO to a single cell performance and even overcome single conventional huge in the energy efficiency and UT per cell.

Future fifth generation (5G) cellular networks have to cope with the expected ten-fold increase in mobile data traffic between 2015 and 2021. To achieve this goal, new technologies are being considered, including massive multiple-input multiple-output (MIMO) systems and millimeter-wave (mmWave) communications. Massive MIMO involves the use of large antenna array sizes at the base station, while mmWave communications employ frequencies between 30 and 300 GHz. In this thesis we study the impact of these technologies on the performance of channel estimators. Our results show that the characteristics of the propagation channel at mmWave frequencies improve the channel estimation performance in comparison with current, low frequency-based, cellular networks. Furthermore, we demonstrate the existence of an optimal angular spread of the multipath clusters, which can be used to maximize the capacity of mmWave networks. We also propose efficient noise variance estimators, which can be employed as an input to existing channel estimators.

In the first part of the dissertation, we provide an extensive overview of sub-6 GHz wireless access technology known as massive multiple-input multiple-output (MIMO) systems, highlighting its benefits, deployment challenges, and the key enabling technologies envisaged for 5G networks. We investigate the fundamental issues that degrade the performance of massive MIMO systems such as pilot contamination, precoding, user scheduling, and signal detection. In the second part, we optimize the performance of the massive MIMO system by proposing several algorithms, system designs, and hardware architectures. To mitigate the effect of pilot contamination, we propose a pilot reuse factor scheme based on the user environment and the number of active users. The results through simulations show that the proposed scheme ensures the system always operates at maximal spectral efficiency and achieves higher throughput. To address the user scheduling problem, we propose two user scheduling algorithms bases upon the measured channel gain. The simulation results show that our proposed user scheduling algorithms achieve better error performance, improve sum capacity and throughput, and guarantee fairness among the users. To address the uplink signal detection challenge in the massive MIMO systems, we propose four algorithms and their system designs. We show through simulations that the proposed algorithms are computationally efficient and can achieve near-optimal bit error rate performance. Additionally, we propose hardware architectures for all the proposed algorithms to identify the required physical components and their interrelationships.

CoordenaÃÃo de AperfeiÃoamento de Pessoal de NÃvel Superior / Este trabalho se propÃe a investigar e propor diferentes tÃcnicas de estimaÃÃo de canal de mÃltiplas entradas e mÃltiplas saÃdas (MIMO) para sistemas de comunicaÃÃo multiusuÃrio operando em regime de interferÃncia em cenÃrio de mÃltiplas cÃlulas. AtenÃÃo particular à dada ao caso onde as estaÃÃes rÃdio-base sÃo equipadas com arranjos de antenas apresentando grande quantidade de antenas, configurando o que se tem referido na literatura como sistemas de comunicaÃÃo MIMO de grande dimensÃo (VL-MIMO, do inglÃs: very large MIMO). Algumas destas tÃcnicas exploram as propriedades das grandes matrizes aleatÃrias e sÃo menos afetadas pela contaminaÃÃo de pilotos. Nesta dissertaÃÃo, os parÃmetros do canal VL-MIMO sÃo estimados a partir de uma decomposiÃÃo em autovalores (EVD, do inglÃs: eigenvalue-decomposition) da matriz de covariÃncia na saÃda do arranjo de antenas receptoras. Esta tÃcnica se mostra menos sensÃvel à presenÃa de interferÃncia do que outras que nÃo exploram propriedades especÃficas da matriz de canal VL-MIMO, como à o caso da soluÃÃo clÃssica dos mÃnimos quadrados (LS, do inglÃs: least-squares). Nesse contexto, propÃe-se ainda uma soluÃÃo para o fator de ambiguidade multiplicativa do mÃtodo baseado em EVD, utilizando um simples produto de Khatri-Rao. Na segunda parte desta dissertaÃÃo, as propriedades dos sistemas VL-MIMO sÃo empregadas num problema de localizaÃÃo de fontes, a fim de determinar a direÃÃo de chegada (DOA) dos sinais incidentes sobre o arranjo, provenientes da cÃlula em questÃo. Explorando o subespaÃo de representaÃÃo dos sinais interferentes, propÃe-se o uso de um algoritmo de classificaÃÃo de tipo MUSIC para estimar a matriz de canal de forma cega. O mÃtodo proposto converte os altos ganhos de resoluÃÃo dos arranjos VL-MIMO em capacidade de reduÃÃo de interferÃncia, podendo fornecer estimativas do canal adequadas, mesmo sob nÃveis fortes de interferÃncia e tambÃm em casos onde os sinais do usuÃrio desejado e dos interferentes sÃo altamente correlacionados espacialmente. Extensas campanhas de simulaÃÃo computacional foram realizadas, dandoum carÃter exploratÃrio a esta dissertaÃÃo no sentido de abranger diferentes cenÃrios e avaliar as tÃcnicas investigadas em comparaÃÃo com soluÃÃes jà consolidadas, permitindo assim a elaboraÃÃo de um panorama mais completo de caracterizaÃÃo dos problemas de estimaÃÃo de parÃmetros no caso VL-MIMO. / The aim of this dissertation is mainly to investigate and propose different channel estimation techniques for a multicell multiuser multiple-input multiple-output (MIMO) communication system. Particular attention is payed to the case that is referred to as very large (VL) MIMO (VL-MIMO) arrays, where the base stations are equipped with a great (or even huge) number of antenna sensors. Some of these techniques exploit properties issued from the (large) Random Matrices Theory and are therefore less affected by the so-called pilot contamination effect. In this work, the parameters of the VL-MIMO channel are estimated from the eigenvalue decomposition (EVD) of the output covariance matrix of the receive antenna array. This technique is more robust to the interference of signals from other cells compared with methods that do not exploit the specific properties of the VL-MIMO channel matrix, which is the case of the classical least squares (LS) solution. In this context, this work also proposes a simpler way to resolve the scaling ambiguity remaining from the EVD-based method using the Khatri-Rao product. The second part of this dissertation exploits the VL-MIMO properties on a source localization problem, aiming to determine the direction of arrival (DoA) of the signals impinging on the antenna array from a given desired cell. Based on the subspace representation of the outer cell interference signals, we propose a new blind MUSIC-like classification algorithm to estimate the channel matrix. The proposed technique convert the high resolution gains of the VL-MIMO arrays into ability to reduce power of undesired signals, yielding good channel estimates even under high interference power levels, and including cases where desired and undesired signals are strongly correlated. Computer simulations have been done in order to cope with different situations and propagation scenarios, thus yielding an exploratory character to our research and allowing us to evaluate and assess the investigated algorithms, comparing them to consolidated solutions in order to establish a complete overview of the parameter estimation problem in the VL-MIMO case.

[pt] Este trabalho tem como finalidade estudar o problema da detecção de usuários no canal reverso de sistemas MIMO de larga escala, que são caracterizados pelo elevado número de elementos de transmissão e recepção, com foco na complexidade computacional e no desempenho em termos de taxa de erro destes sistemas. Inicialmente, os algoritmos de detecção da família Likelihood Ascent Search (LAS) são investigados e é desenvolvido um novo algoritmo de detecção, denominado de Random-List Based LAS (RLBLAS), capaz de atingir melhores taxas de erros com menor complexidade computacional do que os demais detectores considerados. Posteriormente, técnicas de detecção e decodificação iterativas (Iterative Detection and Decoding - IDD) em sistemas MIMO foram analisadas de forma a propor uma estratégia IDD de complexidade computacional reduzida a fim de viabilizar a sua aplicação em cenários massivos. Finalmente, o problema da contaminação por pilotos em sistemas MIMO multicelulares de larga escala, um dos principais limitadores do desempenho desse tipo de sistema, é estudado e estratégias de detecção com cooperação parcial entre as estações base componentes do sistema que visam mitigar os efeitos da contaminação por pilotos são propostas. As análises e afirmações realizadas durante a presente tese são sustentadas por resultados de simulações de Monte Carlo dos sistemas de comunicações em diversos cenários distintos, incluindo os casos em que são considerados os efeitos de correlação entre as antenas de transmissão/recepção, os efeitos de sombreamento e os erros de estimação dos estados dos canais de comunicações envolvidos. / [en] This work focuses on the multi-user multi-cellular large-scale MIMO reverse channel detection problem, where the number of transmitting and receiving antenna elements grows to the order of hundreds. In these scenarios, one major issue is the computational complexity of such systems. Therefore, this thesis aims to propose low-complexity techniques with good BER performance for the reverse channel detection of MIMO systems. Initially, the detection algorithms of the Likelihood Ascent Search (LAS) family are investigated and a new LAS based detector is proposed. This new detector, named Random-List Based LAS (RLB-LAS), is capable of achieving better BER with lower complexity then the other considered detectors. Next, iterative detection and decoding (IDD) techniques are analyzed in order to propose an IDD strategy applied to the detection and decoding of the reverse MIMO channel with reduced complexity to make possible its application to massive scenarios. Finally, the pilot contamination problem in multi-cellular large-scale MIMO systems, one of the major bounds on BER performance of these systems, are studied and some cooperative strategies are proposed in order to reduce the effects of this type of impairments. The analysis and statements of this thesis are supported by Monte Carlo simulation results of the considered systems in different scenarios, including the cases where the effects of transmitting and receiving antenna correlation, log-normal shadowing, and the estimation errors on the channel state information acquisition are considered.