The predominant game-theoretic solutions for distributed rate-maximization algorithms in Gaussian interference channels through optimal power control require perfect channel knowledge, which is not possible in practice due to various reasons, such as estimation errors, feedback quantization and latency between channel estimation and signal transmission. This thesis therefore aims at addressing this issue through the design and analysis of robust gametheoretic algorithms for rate-maximization in Gaussian interference channels in the presence of bounded channel uncertainty. A robust rate-maximization game is formulated for the single-antenna frequency-selective Gaussian interference channel under bounded channel uncertainty. The robust-optimization equilibrium solution for this game is independent of the probability distribution of the channel uncertainty. The existence and uniqueness of the equilibrium are studied and sufficient conditions for the uniqueness of the equilibrium are provided. Distributed algorithms to compute the equilibrium solution are presented and shown to have guaranteed asymptotic convergence when the game has a unique equilibrium. The sum-rate and the price of anarchy at the equilibrium of this game are analyzed for the two-user scenario and shown to improve with increase in channel uncertainty under certain conditions. These results indicate that the robust solution moves closer to a frequency division multiple access (FDMA) solution when uncertainty increases. This leads to a higher sum-rate and a lower price of anarchy for systems where FDMA is globally optimal. A robust rate-maximization game for multi-antenna Gaussian interference channels in the presence of channel uncertainty is also developed along similar principles. It is shown that this robust game is equivalent to the nominal game with modified channel matrices. The robust-optimization equilibrium for this game and a distributed algorithm for its computation are presented and characterized. Sufficient conditions for the uniqueness of the equilibrium and asymptotic convergence of the algorithm are presented. Numerical simulations are used to confirm the behaviour of these algorithms. The analytical and numerical results of this thesis indicate that channel uncertainty is not necessarily detrimental, but can indeed result in improvement of performance of networks in particular situations, where the Nash equilibrium solution is quite inefficient and channel uncertainty leads to reduced greediness of users.

A Wireless Sensor Network (WSN) has wide potential for many applications. It can be employed for normal monitoring applications, for example, the monitoring of environmental conditions such as temperature, humidity, light intensity and pressure. A WSN is deployed in an area to sense these environmental conditions and send information about them to a sink. In certain locations, disasters such as forest fires, floods, volcanic eruptions and earth-quakes can happen in the monitoring area. During the disaster, the events being monitored have the potential to destroy the sensing devices; for example, they can be sunk in a flood, burnt in a fire, damaged in harmful chemicals, and burnt in volcano lava etc. There is an opportunity to exploit the energy of these nodes before they are totally destroyed to save the energy of the other nodes in the safe area. This can prolong WSN lifetime during the critical phase. In order to investigate this idea, this research proposes a new routing protocol called Maximise Unsafe Path (MUP) routing using Ipv6 over Low power Wireless Personal Area Networks (6LoWPAN). The routing protocol aims to exploit the energy of the nodes that are going to be destroyed soon due to the environment, by concentrating packets through these nodes. MUP adapts with the environmental conditions. This is achieved by classifying four different levels of threat based on the sensor reading information and neighbour node condition, and represents this as the node health status, which is included as one parameter in the routing decision. High priority is given to a node in an unsafe condition compared to another node in a safer condition. MUP does not allow packet routing through a node that is almost failed in order to avoid packet loss when the node fails. To avoid the energy wastage caused by selecting a route that requires a higher energy cost to deliver a packet to the sink, MUP always forwards packets through a node that has the minimum total path cost. MUP is designed as an extension of RPL, an Internet Engineering Task Force (IETF) standard routing protocol in a WSN, and is implemented in the Contiki Operating System (OS). The performance of MUP is evaluated using simulations and test-bed experiments. The results demonstrate that MUP provides a longer network lifetime during a critical phase of typically about 20\% when compared to RPL, but with a trade-off lower packet delivery ratio and end-to-end delay performances. This network lifetime improvement is crucial for the WSN to operate for as long as possible to detect and monitor the environment during a critical phase in order to save human life, minimise loss of property and save wildlife.

The rapid growth of wireless services and the breakneck proliferation of wireless devices continue to strain limited spectrum resource. While the need for efficient spectrum sharing mechanisms has been emphasized, opportunistic spectrum access has been considered as a promising mechanism for dynamic spectrum sharing. However, although the idle spectrum could exist, it is usually rather fragmented and distributed, and hence the secondary network users would face the difficulty in finding required contiguous spectrum. Spectrum aggregation can be exploited to provide effective wide bandwidth communication but at the cost of complexity and overhead. When a primary network uses spectrum dynamically, from the nature of opportunistic spectrum access, collisions can occur between primary and secondary transmissions and spectrum handoff can be utilised to provide reliable communication. However, collision occurrence results in spectrum handoff delay in a secondary network user (SU) along with short-term interference to a primary network user (PU). As a SU accesses more spectrum for higher data rates by spectrum aggregation, collisions can occur more frequently and frequent spectrum handoff will be required. While spectrum aggregation will allow the SU to have high flexibility in spectrum use and spectrum handoff can help improve the reliability of secondary transmissions, the SU faces a new spectrum allocation problem: How wide and which parts of spectrum opportunities should be aggregated while considering the complexity and the overhead for aggregation and for spectrum handoff? This thesis addresses the key challenge of opportunistic spectrum access, focusing on efficient spectrum sharing considering the fragmentation of spectrum opportunities in frequency and time domains. First, considering complexity and overhead for aggregation, the spectrum aggregation approach is investigated and guidelines are derived how to reduce spectrum fragmentation for the efficient spectrum utilisation based on simulation results. Second, the relationship between collision occurrence and spectrum aggregation is analysed. Collision probabilities between primary and secondary transmissions are derived and the impacts of spectrum aggregation on data rates and spectrum handoff are investigated. Then, a spectrum aggregation algorithm is proposed to maximise data rates for a given collision probability threshold. Third, when considering spectrum handoff, the impacts of spectrum aggregation on spectrum handoff and short-term interference to PUs are analysed. Then, the spectrum aggregation algorithm is designed with the aim to minimise collision. Finally, the results of this study are summarised, conclusions are presented and a number of future research topics are proposed.

Radio networks are a large part of the business of the Communications Department of Thales UK. The integration, deployment, and configuration of these networks has become a more complex process as the networks have become larger - impacting both the technological development side and the physical network operation side. There is a knowledge gap, most keenly seen when bidding · on large new projects, as to what is possible and what technologies/sub-systems/configurations are best to use in any given system. This sits alongside an imperative to find a way to acquire or generate such knowledge at minimal cost, and certainly without the exorbitant costs of physically building deployment-scale test systems. This thesis addresses that knowledge gap methodologically, via computer simulation models. Simulation of a physical system can generate information that when analysed within the constraints of the model can give useful knowledge about the physical system being simulated. For this reason, a simulation software architecture was designed and implemented, enabling the simulation and modelling of a range of communication networks. This research was done in the Thales RCP (Radio Communications Products) Group, within the Communications Department, and as such its primary use was simulating High Frequency (HF) radios and their operational networks. Models were created of all layers involved in an HF radio system, following the outline of the architecture itself, which is based on the Open Systems Interconnection model. Using this simulation architecture, results gained about the behaviour of HF systems include comparative evaluation studies of HF radio protocols, as well as behaviour and performance profiling of novel protocols and of the configurations of existing systems. This thesis demonstrates that the information obtained from the simulation architecture can be used to address the primary motivating problem: reducing the knowledge gap that exists with these new, large radio network systems. The knowledge gap includes the effects of configuration, protocol, sub-system and so forth on the behaviour and performance of the network amongst other things. Thus, with this kind of knowledge now made available, decision makers are enabled to make more informed decisions at various stages in a project lifecycle, and hence the decision-making process is improved. Aside from internal Thales UK applications, the simulation architecture has been released as open-source. Although the models may be proprietary, the architecture provides an easy-to-use simulation tool, which is capable of quickly developing models for rapid prototyping of any layer in a HF radio communication network system.

This thesis presents an investigation of cooperative transmission techniques and resource block scheduling for maximizing the bandwidth efficiency in the downlink transmission of the urban macro LTE environment. The simulation results show that in order to achieve the highest user bandwidth efficiency normalized by the number of cooperating base stations, over half of the users do not need cooperation, and most of the remaining users choose one of the 2-BS, 3-BS and full cooperation cases to obtain the optimal normalized user bandwidth efficiency. From the simulation results for the resource block scheduling in a small network layout, there are essentially three types of possible optimal cases: the full cooperation case, the full frequency reuse non-cooperative case and finally, the 2/3 reuse non-cooperative case. When shadowing is not included in a 2-cell network and a 3-cell network, over half of the users choose full frequency reuse non-cooperative case as their optimal cases; the handover versions of these three types occur as the fourth type of the optimal cases for a 3-cell network with shadowing effects; the full cooperation case is chosen by over 90% of users in the tri-sectored I-cell network layout. Three sub-optimal algorithms with low complexity are proposed. A location based algorithm can achieve nearly 99% of the optimal total bandwidth efficiency in a non-shadowing environment, and two power based algorithms can realize over 90% of the optimal total bandwidth efficiency in a shadowing environment. Moreover, the second of two power based algorithms can also get approximately 100% of the total bandwidth efficiency for a 3-cell network in a non-shadowing environment and for a tri-sectored 1-cell network layout. For the extended 3-cell network layout, the genetic algorithm is used to obtain a near-optimal solution but without consideration of user fairness. A sub-optimal algorithm with user capacity fairness is proposed. The proposed algorithm can achieve 85% of the total bandwidth efficiency from the genetic algorithm approach and also greatly improves the fairness of the user capacity for the genetic algorithm.

Recent years have seen a rising demand in integrating real world information into existing applications and business processes. The Internet of Things (IoT), which is based on networked embedded devices such as passive and active Radio-frequency identification (RFID) tags, wireless sensor nodes, etc. plays a key role in providing the link to the physical world. Efficient integration of the IoT into Internet of Services is considered as a major challenge, mainly due to the heterogeneity of the existing IoT solutions. The prevalence of IPv6 over Low power Wireless Personal Area Networks (6LoWPAN) has enabled sensor nodes to communicate using Internet technologies and services. While the integration of the Internet Protocol (IP) into wireless sensor networks brings huge opportunities to provide Internet and Web based communications and interactions with other networks, the power, network and resource constraints of the sensor nodes and wireless sensor networks pose major challenges to adopting IP-based solutions in wireless sensor networks. Service Oriented is a well-established architecture and has been successfully used for many years on the Internet. Service Oriented Architecture (SOA) and the web services provide interoperability and reusability of software components and enabling the design of distributed and loosely coupled systems that are supported by multiple platforms. Using service oriented interactions over wireless sensor networks can enable seamless integration of the data and functionalities that are provided by these networks. Using service oriented architecture, the data and functionalities of wireless sensor nodes and networks can be represented as a service to the Internet and Web. The services can also be used for managing and monitoring the networks. This will create a huge potential for interaction and integration of the wireless sensor network enabled services into the Web. This thesis describes the design and development of a flexible and efficient Web service Architecture for wireless sensor networks. The research has developed an energy efficient and adaptable discovery mechanism for ad hoc and dynamic wireless sensor networks. Our proposed solution is based on the IP protocol suite and takes into account the inherent resource limitation of wireless sensor networks. Our solutions support both SOAP and RESTful services in an integrated framework and can interact with different types of services on the Web and enterprise systems. We also provide solutions for service discovery that uses a reinforcement learning technique to improve the performance of the service discovery process according to the dynamicity of the network. The work is implemented and evaluated on a wireless sensor test-bed accompanied by a demonstrator application to represent the functionality of the proposed solutions. The evaluation results show that the proposed solutions can significantly reduce the payload and size of the service discovery message in service communications up to 94% compared with the traditional service oriented solutions over the Web. The proposed service discovery mechanisms improve the power and communication efficiency by decreasing size of the service discovery messages and the process overload by up to 25%. The proposed solutions can co-operate with other solutions such as Constrained Application Protocol (CoAP); however the discovery and service communication solutions significantly reduce the messaging overload and can enhance CoAP or other SOAP based services. The discovery process can also be used with SOAP and CoAP (i.e. RESTful) services and enables efficient discovery solutions for the services in dynamic wireless sensor networks.

In this thesis new methods are presented for achieving spectrum sensing in cognitive radio wireless networks. In particular, supervised, semi-supervised and unsupervised machine learning based spectrum sensing algorithms are developed and various techniques to improve their performance are described. Spectrum sensing problem in multi-antenna cognitive radio networks is considered and a novel eigenvalue based feature is proposed which has the capability to enhance the performance of support vector machines algorithms for signal classification. Furthermore, spectrum sensing under multiple primary users condition is studied and a new re-formulation of the sensing task as a multiple class signal detection problem where each class embeds one or more states is presented. Moreover, the error correcting output codes based multi-class support vector machines algorithms is proposed and investigated for solving the multiple class signal detection problem using two different coding strategies. In addition, the performance of parametric classifiers for spectrum sensing under slow fading channel is studied. To address the attendant performance degradation problem, a Kalman filter based channel estimation technique is proposed for tracking the temporally correlated slow fading channel and updating the decision boundary of the classifiers in real time. Simulation studies are included to assess the performance of the proposed schemes. Finally, techniques for improving the quality of the learning features and improving the detection accuracy of sensing algorithms are studied and a novel beamforming based pre-processing technique is presented for feature realization in multi-antenna cognitive radio systems. Furthermore, using the beamformer derived features, new algorithms are developed for multiple hypothesis testing facilitating joint spatio-temporal spectrum sensing. The key performance metrics of the classifiers are evaluated to demonstrate the superiority of the proposed methods in comparison with previously proposed alternatives.

Efficient estimation and processing of high-dimensional data is important in many scientic and engineering domains. In this thesis, we explore structured sensing methods for high-dimensional signal in three different perspectives: structured random matrices for compressed sensing and corrupted sensing, atomic norm regularization for massive multiple-input-multiple-output (MIMO) systems and variable density sampling for random field. Designing efficient sensing systems for high-dimensional data by appealing to the prior knowledge that their intrinsic information is usually small has become popular in recent years. As a starting point, compressed sensing has proven to be feasible for estimating sparse signals when the number of measurements is far less than the dimensionality of the signals. Besides fully random sensing matrices, many structured sensing matrices have been designed to reduce the computation and storage cost. We propose a unified structured sensing framework and prove the associated restricted isometry property. We demonstrate that the proposed framework encompasses many existing designs. In addition, we construct new structured sensing models based on the proposed framework. Furthermore, we consider a generalized problem where the compressive measurements are affected by both dense noise and sparse corruption. We show that in some cases the proposed framework can still guarantee faithful recovery for both the sparse signal and the corruption. The next part of the thesis is concerned with channel estimation and faulty antennas detection in massive MIMO systems. By leveraging the intrinsic information of the channel matrix through atomic norm, we propose new algorithms and demonstrate their performances for both channel estimation and faulty antennas detection. In the last part, we propose a variable density sampling method for the estimation of high-dimensional random field. While conventional uniform sampling requires a number of samples increasing exponentially with the dimension, we show that faithful recovery can be guaranteed with a polynomial size of random samples.

A fundamental problem in communications is the estimation of the channel. The signal transmitted through a communications channel undergoes distortions so that it is often received in an unrecognizable form at the receiver. The receiver must expend significant signal processing effort in order to be able to decode the transmit signal from this received signal. This signal processing requires knowledge of how the channel distorts the transmit signal, i.e. channel knowledge. To maintain a reliable link, the channel must be estimated and tracked by the receiver. The estimation of the channel at the receiver often proceeds by transmission of a signal called the 'pilot' which is known a priori to the receiver. The receiver forms its estimate of the transmitted signal based on how this known signal is distorted by the channel, i.e. it estimates the channel from the received signal and the pilot. This design of the pilot is a function of the modulation, the type of training and the channel.

Gravity waves play a crucial role in the dynamics of the middle atmosphere through the vertical transport of tropospheric energy and momentum. Despite the importance of gravity wave effects, nearly all general circulation models significantly underestimate gravity wave drag around the southern wintertime polar vortex, leading to large discrepancies from observed behaviour and a strong impediment to model progress. Here we use GPS radio occultation data from the COSMIC satellite constellation to investigate key properties of gravity waves in and around the southern polar vortex and over the gravity wave hot spot of the southern Andes/Drake Passage/Antarctic Peninsula. We also develop spectral analysis tools in order to further the capabilities of our data for this purpose. By analysing vertical profiles of atmospheric temperature from COSMIC, we find evidence of the meridional propagation of waves into the polar vortex from sources far to the north. We develop a new wavelet-based analysis technique for the quantitative identification of gravity waves in COSMIC profiles, and use it to investigate gravity wave intermittency over the hot spot region and around the edge of the polar vortex. We then estimate gravity wave momentum flux over the hot spot from closely-spaced pairs of COSMIC profiles. Finally, we develop a new two-dimensional spectral analysis method for the measurement of gravity wave amplitudes, horizontal wavelengths and directions of propagation from AIRS measurements, based upon the two-dimensional Stockwell transform. We show that, by using an alternative elliptical spectral window, we can dramatically improve the measurement of wave amplitude. We apply our two-dimensional Stockwell transform to AIRS measurements over the known gravity wave hot spots of South Georgia and the Antarctic Peninsula, measuring gravity properties and momentum fluxes with improved confidence, accuracy and rigour over current methods.