Cloud gaming over mobile networks enables players to play high-resource consuming games on low-end devices with various intrinsic restrictions such as limited battery lifetime and computational capacity. For mobile cloud gaming(MCG), the remaining battery level on the device is one of the critical factors that affect the sensitivity of user satisfaction. Thus, an android application is developed to measure the energy consumption of mobile devices that measure the power consumption of the device such that the obtained values correspond with the specific network conditions and users. The collected values are studied to identify if the energy consumption of the device is impacted by the network degradation that might occur during MCG in cellular networks. Results demonstrate that the energy consumption is at its highest when packet loss is 45% at 2ms RoundTrip Time (RTT) delay. Moreover, a qualitative study on the perceived Quality of Experience (QoE) of MCG over mobile networks is conducted and its impact on the energy consumption of the device is investigated where 31 users play a cloud-based First Person Shooter (FPS) for approximately 2 hours each. The results demonstrate the existence of the relationship between energy consumption and perceived QoE whereas negates the hypothesis of the existence of the relationship between QoE and CPU resources. In addition, to make comparisons of energy consumption of MCG with online mobile gaming (OMG), another test is carried out where each user plays another non-cloud-based FPS game and it is found that MCG is more efficient than OMG under the least energy-consuming network condition (2ms RTT delay) by 33.3% and the most energy consuming network condition (45% packet loss at 2ms RTT) by 32.7% in 4G cellular network.

During the thesis project, an evaluation tool was developed for Telenor. This tool identifies the most reliable cell within a site based on its standard deviation and systematically ranks the performance of each cell using key performance indicators (KPIs). The tool also calculates the mean and the median to allow the user to get an overview over the network performance. The analysis underscored the importance of robust network reliability, especially when considering the deployment of home care technologies in remote areas. The tool is designed to analyse extracted data, providing Telenor with a view of network performance and ensuring top-tier service quality in even the most remote and challenging terrains. / Under examensarbetet lades fokus på skapandet av ett utvärderingsverktyg för Telenor. Detta verktyg bestämmer den mest tillförlitliga cellen inom en basstation baserat på dess standardavvikelse för den analyserade KPI och klassificerar varje cells prestanda. Utöver det räknas medelvärdet och medianen ut för att användaren ska få en bättre översikt över nätverksprestandan. Utbyggnaden av hemvårdsteknologier i avlägsna områden tas särskilt i beaktande. Verktyget är designat för analys av data och ger en översikt över nätverkets prestanda, vilket kan bidra till att säkerställa optimal servicekvalitet.

大型天然災害會癱瘓通訊系統，嚴重影響到救災效率，本論文旨在快速進行可用連外頻寬分配，供應急通訊系統使用。無線通訊技術的成熟，為使用者帶來極大的便利性，但當發生大規模的地震或強烈颱風等重大天然災害時，通訊系統卻常常因架構等因素，隨著電力與交通系統的損毀而癱瘓。由歷年大型災變中多數災區內之行動通訊系統全面中斷即可印證行動通訊系統其實是極為脆弱，而有效運作的通訊系統卻是災情傳遞、資源調度以及互助協調是否順利的關鍵因素。 本篇論文所探討的應急通訊系統是利用倖存的連通基地台和斷訊卻沒有損毀的基地台，以無線電連接起來建構一個臨時性的通訊系統，稱為應急蜂巢式行動通訊網路(Contingency Cellular Network，CCN)。由於CCN的連外頻寬有限，大量話務將造成通訊系統壅塞，影響重要訊息傳遞，且災區各個地方受災情況不盡相同，使得 CCN 的頻寬資源需視各地災情緊急程度與需求進行規劃配置，以充分發揮頻寬效益傳遞重要資訊。本論文主要在探討如何在CCN網路拓樸已決定的情況下進行頻寬分配，以達到最大的救災效益。因此我們提出一適合 CCN 樹狀結構的頻寬分配優化模型，以追求救災效益的最大化，這個模型可供使用者(救災指揮單位)系統化的解決 CCN 頻寬分配問題。 本論文所提出的頻寬分配模型包含 CCN 樹狀拓樸、基地台數目、可用之連外頻寬資源限制、各基地台Backhaul頻寬限制、基本頻寬需求限制、差異化之通訊品質通道和效益遞減函數。我們證明此模型是NP-Hard問題，並提出一個考慮各基地台的災情緊急程度以及通訊品質需求差異而進行快速頻寬分配的演算法，此演算法透過計算頻寬分配總救災效益決定優劣。經實驗，可快速得出接近最佳解的頻寬分配結果。 / When stricken by a large-scale disaster, the efficiency of disaster response operation is very critical to life saving. We propose to build a contingency cellular network to support emergency communication in large scale natural disasters by connecting disconnected base stations. This thesis addresses the bandwidth allocation problem. The advance of mobile communication technologies has brought great convenience to users. Cellular phone becomes the first communication tool most people would use in emergency. However, cellular networks were usually crashed in earthquake, typhoons or other natural disasters due to power outage or backhaul breakage. Unfortunately, the efficiency of communication system is a critical factor to the success of disaster response operation such as resource allocation as well as coordination of rescue and relief operations. We designed a contingency cellular network (CCN) by connecting physically intact but service-disrupted base stations together with wireless links. As the bandwidth resource in CCN is limited, a smart bandwidth allocation to facilitate prioritized bandwidth sharing will maximize the contribution of CCN to the disaster response operation. We model the CCN Bandwidth Allocation Problem into a Nested 0-1 Knapsack Problem aiming to maximize disaster operation efficiency. The problem is proven to be NP Hard. We also design an efficient heuristic algorithm to solve the problem when it is needed in urgent.

Cette thèse s'intéresse aux problèmes d'allocations des ressources et de puissance dans les réseaux cellulaires de quatrième génération (4G). Pour faire face à la demande continuellement croissante en débit des utilisateurs mobiles, les opérateurs n'ont d'autre choix que de densifier leurs infrastructures d'accès au réseau radio (RAN), afin de maximiser l'utilisation de la bande passante disponible dans l'espace. Un des défis de cette nouvelle architecture est la coexistence de nombreuses cellules voisines et la gestion des interférences co-canal qu'elles génèrent entre elles. De telles contraintes ont amené la communauté scientifique à s'intéresser aux réseaux auto-organisés et auto-optimisés (SON), qui permettent aux réseaux de s'optimiser localement via des décisions décentralisées (sans planification statique). L'intérêt principal de tels réseaux est le passage à l'échelle des algorithmes distribués et la possibilité de s'adapter dynamiquement à de nouveaux environnements. Dans cette optique, nous proposons l'étude de deux problèmes d'allocation de ressources. La première partie de cette thèse se concentre sur l'optimisation de l'usage des ressources, dans un contexte de transmission coordonnée par plusieurs stations de base (CoMP). Les performances de la coordination de stations de base sont évaluées, selon le critère de capacité uniforme, ainsi que le compromis entre l'efficacité spectrale et l'équité entre les utilisateurs. Nous proposons également une méthode généralisée et distribuée de sélection de l'ensemble de stations en coopération, afin d'optimiser le compromis efficacité-équité. Dans une seconde partie, nous nous intéressons à l'optimisation de l'allocation des ressources et de puissance, dans le but de minimiser la consommation électrique du réseau. Nous présentons deux algorithmes dont les décisions sont décentralisées. Le premier est basé sur une optimisation stochastique (via l'échantillonneur de Gibbs) et permet une optimisation globale du système. Le second quant à lui est basé sur l'adaptation de la théorie du contrôle et utilise des modèles prédictifs et la poursuite de cibles pour allouer les ressources et les puissances dans un contexte de canaux et d'interférences dynamiques. Dans de nombreux cas, plusieurs objectifs concurrents sont à considérer pour évaluer les performances d'un réseau (capacité totale, équité, consommation électrique, etc.). Dans le cadre de cette thèse, nous nous efforçons à présenter les résultats sous la forme de compromis multi-objectifs. / This thesis focuses on resources and power allocation problem in the fourth generation (4G) of cellular networks. To face the continuous growth of mobile users capacity requirements, operators need to densify their radio access network (RAN) infrastructure, to maximize the use of the available bandwidth in space. One of the major issues of this new architecture is the proximity of many base stations (BS) and the management of the interference they generate on each other's cell. Such constraints makes scientific community focus on Self-Optimized, Self-Organized Networks (SON) that allow network elements to optimize them-selves through decentralized decisions (no static network planning is required). A major interest of SON is their capability to scale to large and non-organized networks, as well as being able to adapt them-selves dynamically, by using distributed algorithms. In this context, this thesis proposes the study of two resource allocation problems. The first part of this thesis focuses on the optimisation of resource sharing, in the context of coordinated multi-points transmissions (CoMP). Performances of BS coordination are evaluated, using the uniform capacity criterion, as well as the trade-off between total capacity and fairness among users. We also propose a generalized and distributed method to select the set of coordination of BS, to optimize the capacity-fairness trade-off. In the second part of this thesis, we focus on optimizing the transmit power and resource allocation, in order to reduce electric consumption. We present two distributed algorithms: the first one is based on a stochastic optimisation (using Gibbs sampling), and tries to reach the global optimum state through decentralized decision. The second one is based on control theory, and uses target tracking as well as model predictive control to allocate resources and power in a dynamic channel scenario. In many cases, trade-offs are to be maid between opposite objectives when evaluating network performances (total throughput, fairness, energy consumption, etc.). In this thesis, we present most of the network performances using multi-objectives evaluations.

[pt] Na primeira parte desta tese, são desenvolvidos dois esquemas de detecção por listas para sistemas MIMO multiusuário. As técnicas propostas usam uma única transformação de redução de reticulado (LR) para modificar a matriz de canal entre os usuários e a estação base (BS). Após a transformação LR, um candidato confiável do sinal transmitido é obtido usando um detector de cancelamento sucessivo de interferências (SIC). No detector em múltiplos ramos com redução de reticulado e cancelamento sucessivo de interferências (MB-LR-SIC) proposto, um número fixo de diferentes ordenamentos para o detector SIC gera uma lista de possíveis candidatos para a informação transmitida. O melhor candidato é escolhido usando o critério maximum likelihood (ML). No detector por listas de tamanho variável (VLD) proposto, um algoritmo que decide se o candidato atual tem uma boa qualidade ou se é necessário continuar procurando por um candidato melhor nos ordenamentos restantes é utilizado. Os resultados numéricos mostram que os esquemas propostos têm um desempenho quase ótimo com uma complexidade computacional bem abaixo do detector ML. Um esquema de detecção e decodificação iterativa (IDD) baseado no algoritmo VLD é também desenvolvido, produzindo um desempenho próximo a um sistema mono usuário (SU) livre de interferências. Na segunda parte desta tese, uma técnica de detecção desacoplada de sinais (DSD) para sistemas MIMO massivo é proposta. Esta técnica permite que o sinal composto recebido na BS seja separado em sinais independentes, correspondentes a diferentes classes de usuários, viabilizando assim o uso dos procedimentos de detecção propostos na primeira parte desta tese em sistemas MIMO massivos. Um modelo de sinais para sistemas MIMO massivo com antenas centralizadas e/ou antenas distribuídas operando em redes heterogêneas de quinta geração é proposto. Uma análise baseada na soma das taxas e um estudo de custo computacional para DSD são apresentados. Os resultados numéricos ilustram o excelente compromisso desempenho versus complexidade obtido com a técnica DSD quando comparada com o esquema de detecção conjunta tradicional. / [en] In the first part of this thesis, we introduce two list detection schemes for the uplink scenario of multiuser multiple-input multiple-output (MUMIMO) systems. The proposed techniques employ a single lattice reduction (LR) transformation to modify the channel matrix between the users and the base station (BS). After the LR transformation, a reliable candidate for the transmitted signal vector, provided by successive interference cancellation (SIC) detection is obtained. In the proposed multi-branch lattice reduction SIC (MB-LR-SIC) detector, a fixed number of different orderings, generates a list of SIC detection candidates. The best candidate is chosen according to the maximum likelihood (ML) selection criterion. For the proposed variable list detection (VLD) scheme, an algorithm to decide if the current candidate has good quality or if it is necessary to further explore different orderings to improve the detection performance is employed. Simulation results indicate that the proposed schemes have a near-optimal performance while keeping its computational complexity well below that of the ML detector. An iterative detection and decoding (IDD) scheme based on the VLD algorithm is also developed, producing an excellent performance that approaches the single user (SU) scenario. In the second part of this thesis, a decoupled signal detection (DSD) technique which allows the separation of uplink signals, for each user class, at the base station (BS) for massive MIMO systems is proposed. The proposed DSD allows to implement the detection procedures proposed in the first part of this thesis in massive MIMO scenarios. A mathematical signal model for massive MIMO systems with centralized and distributed antennas in the future fifth generation (5G) heterogeneous cellular networks is also developed. A sum-rate analysis and a study of computational cost for DSD are also presented. Simulation results show excellent performance of the proposed DSD algorithm when combined with linear and SIC-based detectors.

Cellular network traffic prediction is a critical challenge for communication providers, which is important for use cases such as traffic steering and base station resources management. Traditional prediction methods mostly rely on historical time-series data to predict traffic load, which often fail to model the real world and capture surrounding environment conditions. In this work, we propose a multi-modal deep learning model for 4G/5G Cellular Network Traffic prediction by considering external data sources such as satellite imagery and weather data. Specifically, our proposed model consists of three components (1) temporal component (modeling correlations between traffic load values with historical data points via LSTM) (2) computer vision component (using embeddings to capture correlations between geographic regions that share similar landscape patterns using satellite imagery data and state of the art CNN models), and (3) weather component (modeling correlations between weather measurements and traffic patterns). Furthermore, we study the effects and limitations of using such contextual datasets on time series learning process. Our experiments show that such hybrid models do not always lead to better performance, and LSTM model is capable of modeling complex sequential interactions. However, there is a potential for classifying or labelling regions by their urban landscape and the network traffic. / Förutsägelse av mobilnätstrafik är en kritisk utmaning för kommunikation leverantörer, där användningsområden inkluderar trafikstyrning och hantering av basstationsresurser. Traditionella förutsägelsesmetoder förlitar sig främst på historisk tidsseriedata för att förutsäga trafikbelastning, detta misslyckas ofta med att modellera den verkliga världen och fånga omgivande miljö. Det här arbetet föreslår en multimodal modell med djupinlärning förutsägelse av 4G/5G nätverkstrafik genom att beakta externa datakällor som satellitbilder och väderdata. Specifikt består vår föreslagna modell av tre komponenter (1) temporal komponent (korrelationsmodellering mellan trafikbelastningsvärden med historiska datapunkter via LSTM) (2) datorseende komponent (med inbäddningar för att fånga korrelationer mellan geografiska regioner som delar liknande landskapsmönster med hjälp av satelitbilddata och state-of-the-art CNN modeller), och (3) väderkomponent (modellerande korrelationer mellan vädermätningar och trafikmönster). Dessutom studerar vi effekterna och begränsningarna av att använda sådana kontextuella datamängder på tidsserieinlärningsprocessen. Våra experiment visar att hybridmodeller inte alltid leder till bättre prestanda och att LSTM-modellen är kapabel att modellera komplexa sekventiella interaktioner. Det finns dock en potential att klassificera eller märka regioner efter deras stadslandskap och nättrafiken. / La prévision du trafic sur les réseaux cellulaires est un défi crucial pour les fournisseurs de communication, ce qui est important pour les cas d’utilisation tels que la direction du trafic et la gestion des ressources des stations de base. Les méthodes de prédiction traditionnelles reposent principalement sur des données historiques de séries chronologiques pour prédire la charge de trafic, qui échouent souvent à modéliser le monde réel et à capturer les conditions de l’environnement environnant. Dans ce travail, nous proposons un modèle d’apprentissage profond multimodal pour la prédiction du trafic des réseaux cellulaires 4G/5G en considérant des sources de données externes telles que l’imagerie satellitaire et les données météorologiques. Plus précisément, notre modèle proposé se compose de trois composants (1) composant temporel (modélisation des corrélations entre les valeurs de charge de trafic avec des points de données historiques via LSTM) (2) composant de vision par ordinateur (utilisant des incorporations pour capturer les corrélations entre les régions géographiques qui partagent des modèles de paysage similaires à l’aide de données d’imagerie satellitaire et de modèles CNN de pointe) et (3) composante météorologique (modélisation des corrélations entre les mesures météorologiques et les modèles de trafic). De plus, nous étudions les effets et les limites de l’utilisation de tels ensembles de données contextuelles sur le processus d’apprentissage des séries chronologiques. Nos expériences montrent que de tels modèles hybrides ne conduisent pas toujours à de meilleures performances, et le modèle LSTM est capable de modéliser des interactions séquentielles complexes. Cependant, il est possible de classer ou d’étiqueter les régions en fonction de leur paysage urbain et du trafic du réseau.

[pt] Com o advento dos sistemas celulares de segunda e terceira gerações é esperado que as operadoras se vejam obrigadas a enfrentar um aumento dramático na carga de sinalização que trafega sobre a parte fixa da rede móvel. Apesar disso, são raros os provedores de serviços de comunicações móveis que possuem atualmente a capacidade de prever com relativa precisão o montante desse aumento. Este trabalho apresenta as etapas do desenvolvimento de uma ferramenta de simulação para análise de desempenho de redes de comunicação móvel celular com base na carga de sinalização gerada pelos procedimentos que a mantém em operação. A plataforma de simulação inclui um modelo de mobilidade e teletráfego para caracterizar o processo de geração dos cenários típicos de uma rede móvel celular e um modelo de retardos para representação da rede de sinalização. Ao final do estudo,são apresentados exemplos de aplicação da ferramenta na obtenção de resultados sobre gerência de status, gerência de localização, avaliação da carga de sinalização,dimensionamento da rede de sinalização e análise de desempenho para diferentes configurações de rede. / [en] The advent of second and third generation cellular systems make cellular operators face dramatic increase in the signaling traffic over the fixed part of the mobile network. In spite of this, rare mobile communications service providers are able to forecast the above mention increase and quantify it with reasonable precision. This work describes the development process of a simulation tool for performance analysis of cellular mobile network based on the signaling load generated by the procedures that keeps it working. The simulation platform inlcudes a mobility and teletraffic model to describe the generation process of cellular mobile networks tipical scenarios and a delay model to represent the signaling network. At the end, examples showing the application of the simulation tool to obtain results about status and location management, signaling load evaluation, signaling network planning and performance analysis for different network configurations are presented.