Resource Management In Wireless Networks And Big Data Analysis Based On Distributed ADMM Algorithms

With the arrival of the big data era and the popularization of various intelligent devices as well as the Internet,the data amount that is generated daily is increasing explosively.As a result,the information gathering,communication and analysis regarding big data are facing brand new challenges.How to efficiently acquire,communicate and analyze these data to maximize their potential value will become the key factors that can influence the future developing directions of the world's economics,science and technology.In this thesis,we focus on the current issues in data sampling,data communication and data analysis techniques under the big data background.In particular,we aim at solving the resource management problem in wireless networks,the mobile data offloading problem,the resource allocation problem based on distributed optimization algorithms and the prediction of air quality problem based on big data mining.We summarize our main contributions as follows.1).Regarding the big data communication problem,we propose a distributed decompositioncoordination framework for mobile data offloading in wireless communication networks.Mobile data offloading allows alternative network systems such as Wi-Fi hotspot access points to offload the traffic that is originally targeted for the cellular network operators.It can alleviate the cellular network congestions and enhance users' quality-of-service.One of the main challenges of the mobile data offloading is to develop simple and distributed traffic allocation strategies that can optimally allocate the cellular data traffic from multiple cellular network operators to the access points.In particular,we propose a distributed optimization framework with decomposition-coordination to solve this problem.In our framework,the utility maximization problem of the cellular network system with mobile data offloading is formulated as a non-smooth convex optimization problem.We then divide this problem into a set of sub-problems,and each of them can be solved independently.All the sub-problems are coordinated with each other by a virtual data offloading coordinator which can collect the intermediate calculation values from cellular network operators and access points.The coordinator will then feedback the coordination results calculated from the collected values.We propose two distributed algorithms that can achieve the global optimization solution under different scenarios.The first one is the multi-block proximal Jacobi alternating direction method of multipliers.In this algorithm,the coordination is assumed to be fully synchronized,and the coordinator will only feedback the coordination result after successfully re-ceiving all the values from the subproblems.We relax this assumption by introducing the second algorithm referred to as the distributed asynchronized alternating direction method of multipliers.In this algorithm,the coordination does not need to be perfectly synchronized and the coordinator will feedback a coordination result whenever it receives a value.We prove that both algorithm can achieve the global optimal solution.Numerical results under various network settings and conditions show that the proposed framework and algorithms can significantly improve the cellular network operators' revenue and mobile users' quality-of-service.2).Regarding the big data sampling problem,we propose a hybrid data offloading framework for energy efficient wireless sensor networks.With the fast growing demand for Internet-of-Things-based smart systems and intelligent cities,modern wireless sensor networks are required to support high-speed data transmission with very low energy consumption.How to develop a distributed framework for wireless sensor networks that can reduce the energy consumption and maintain the quality-of-service for the information collection and communication service is still an open problem.This thesis considers the wireless sensor network in an urban environment co-located with multiple cellular small cells and the Wi-Fi systems.We propose a new framework referred to as the hybrid data offloading.In this framework,instead of directly sending data to the fusion center,each sensor node can send its collected data by utilizing the co-located alternative wireless technologies.In particular,the wireless sensor network can offload its traffic using three possible strategies:offloading traffic to neighboring Wi-Fi access points,sending data via the licensed band rented from cellular service providers and offloading traffic to cellular small-cell base stations.We develop a distributed algorithm based on distributed alternating direction method of multipliers with asynchronous coordination which allows each sensor node to decide the optimal traffic amount sent by each of the above strategies.We prove that our proposed algorithm converges to the global optimal solution in linear time.Numerical results show that if the number of co-existing Wi-Fi devices is small(less than 5),our proposed hybrid data offloading framework can save around 75% of the energy consumed by the sensor network compared to other offloading strategies.3).Regarding the big data analysis problem,we propose a high resolution and accuracy air pollutant concentration prediction method from heterogeneous big data resources.High quality air pollutant concentration information is of vital importance to the study of the for-mation of air pollution and how to reduce its damage on human health.One of the traditional approaches to monitor air quality is using static ground monitor stations.Due to the high costs in both financial and human resources to build and maintain these monitor stations,the resolution of the obtained air pollutant concentration information is poor.To solve this problem,various approaches,such as regression and classification approaches etc.,have been proposed to predict high resolution spatio-temporal air pollutant concentration map.However,it is still difficult for the prediction accuracy of these approaches to meet the requirements in many practical situations.In this thesis,we take advantage of heterogeneous big data sources to reconstruct a high resolution spatial-temporal air pollutant concentration cube.Firstly,we predict a preliminary high resolution air pollutant concentration cube from measurements of both ground monitor stations and mobile stations equipped with sensors,as well as various meteorology and geography covariates.Our model is based on the stochastic partial differential equations approach and we use the integrated nested Laplace approximation algorithm as an alternative to the Markov Chain Monte Carlo methods to improve the computational efficiency.Next,in order to further improve the accuracy of the predicted concentration cube,we model the issue as a convex and sparse optimization problem.In particular,we minimize the total variant along with constraints involving satellite observed low resolution air pollutant data and the aforementioned measurements from ground monitor stations and mobile platforms.We propose two algorithms to solve the total variation minimization problem.Numerical simulations on real data sets in Piemonte area in Italy show significant improvements in the accuracy of the reconstructed air pollutant concentration cube.