Following the rapid development of mobile infrastructure and hand-held user devices like smartphones and tablets, the way that mobile networks have been utilized has changed accordingly. Users are no longer using their phones just to talk to and message each other, but have migrated most of their multimedia consumption to these devices. The increased network strain and reduced profit margins that new services bring to the table might lead to future decoupling between revenues and expenses in the network. In order to avoid this scenario various different strategies have been proposed and one of them is introduction of QoS/QoE related resource allocation techniques. This thesis looks at Interruption Prediction scheduler which tries to predict and prevent audio or video interruptions experienced during video conference call. In order to test the performance of these schedulers, MATLAB simulator reusing some of the functions from RUNE toolbox has been used. Results of the newly presented algorithms are compared with a QoE/QoS agnostic scheduler. Unfortunately results of these schedulers in the case of video conference are not as promising as they were in the case of some other types of services. The research done shows that these modifications of the scheduling algorithms do not help to increase the network performance in terms of observed KPIs (Total Time of Interruptions, Frequency of Interruptions and network throughput) in most of the described scenarios. However, there are some special scenarios when these schedulers show a certain potential.

A new era of wireless networks is evolving, thanks to the significant advances in communications and networking technologies. In parallel, wireless services are witnessing a tremendous change due to increasingly heterogeneous and stringent demands, whose quality of service requirements are expanding in several dimensions, putting pressure on mobile networks. Examples of those services are augmented and virtual reality, as well as self-driving cars. Furthermore, many physical systems are witnessing a dramatic shift into autonomy by enabling the devices of those systems to communicate and transfer control and data information among themselves. Examples of those systems are microgrids, vehicles, etc. As such, the mobile network indeed requires a revolutionary shift in the way radio resources are assigned to those services, i.e., RRM. In RRM, radio resources such as spectrum and power are assigned to users of the network according to various metrics such as throughput, latency, and reliability. Several methods have been adopted for RRM such as optimization-based methods, heuristics and so on. However, these methods are facing several challenges such as complexity, scalability, optimality, ability to learn dynamic environments. In particular, a common problem in conventional RRM methods is the failure to adapt to the changing situations. For example, optimization-based methods perform well under static network conditions, where an optimal solution is obtained for a snapshot of the network. This leads to higher complexity as the network is required to solve the optimization at every time slot. Machine learning constitutes a promising tool for RRM with the aim to address the conflicting objectives, i.e., KPIs, complexity, scalability, etc. In this thesis, we study the use of reinforcement learning and its derivatives for improving network KPIs. We highlight the advantages of each reinforcement learning method under the studied network scenarios. In addition, we highlight the gains and trade-offs among the proposed learning techniques as well as the baseline methods that rely on either optimization or heuristics. Finally, we present the challenges facing the application of reinforcement learning to wireless networks and propose some future directions and open problems toward an autonomous wireless network. The contributions of this thesis can be summarized as follows. First, reinforcement learning methods, and in particular model-free Q-learning, experience large convergence time due to the large state-action space. As such, deep reinforcement learning was employed to improve generalization and speed up the convergence. Second, the design of the state and reward functions impact the performance of the wireless network. Despite the simplicity of this observation, it turns out to be a key one for designing autonomous wireless systems. In particular, in order to facilitate autonomy, agents need to have the ability to learn/adjust their goals. In this thesis, we propose transfer in reinforcement learning to address this point, where knowledge is transferred between expert and learner agents with simple and complex tasks, respectively. As such, the learner agent aims to learn a more complex task using the knowledge transferred from an expert performing a simpler (partial) task.

The third generation of mobile communication systems introduce interactive Multicast and Unicast multimedia services at a fast data rate of up to 2 Mbps and is expected to complete the globalization of the mobile telecommunication systems. The implementation of these services on satellite systems, particularly for broadcast and multicast applications to complement terrestrial services is ideal since satellite systems are capable of providing global coverage in areas not served by terrestrial telecommunication services. However, the main bottleneck of such systems is the scarcity of radio resources for supporting multimedia applications which has resulted in the rapid growth in research efforts for deriving efficient radio resource management techniques. This issue is addressed in this thesis, where the main emphasis is to design a dynamic scheduling framework and algorithm that can improve the overall performance of the radio resource management strategy of a UMTS compatible satellite network, taking into account the unique characteristics of wireless channel conditions. This thesis will initially be focused on the design of the network and functional architecture of a UMTS -compatible satellite network. Based on this architecture, an effective scheduling framework is designed, which can provide different types of resource assigning strategies. A functional model of scheduler is defined to describe the behaviours and interactions between different functional entities. An OPNET simulation model with a complete network protocol stack is developed to validate the performance of the scheduling algorithms implemented in the satellite network. Different types of traffic are considered for the OPNET simulation, such as the Poisson Process, ONOFF Source and Self Similar Process, so that the performance of scheduling algorithm can be analyzed for different types of services. A novel scheduling algorithm is proposed to optimise the channel utilisation by considering the characteristics of the wireless channel, which are bursty and location dependent. In order to overcome the channel errors, different code rates are applied for the user under different channel conditions. The proposed scheduling algorithm is designed to give higher priority to users with higher code rate, so that the throughput of network is optimized and at the same time, maintaining the end users¿ service level agreements. The fairness of the proposed scheduling algorithm is validated using OPNET simulation. The simulation results show that the algorithm can fairly allocate resource to different connections not only among different service classes but also within the same service class depending on their QoS attributes. / Inmarsat Global Ltd. BGAN and the European Space Agency (ESA)

With the increasing demand for wireless communications, radio resource management (RRM) plays an important role in future wireless networks in order to provide higher data rates and better quality of services, given the limited amount of available radio resources. Although some specific features of wireless communication networks cause challenges to effective and efficient RRM, they bring opportunities that help improv- ing network performance and resource utilization. In this thesis, we focus on RRM issues related to the broadcast/multicast nature in wireless communication networks. The work is divided into two parts. In the first part, we exploit how to take advantage of the broadcast nature of wire- less transmissions in RRM by opportunistically applying two-way relaying (or network coding) and traditional one-way relaying. Different objectives are considered, includ- ing maximizing total packet transmission throughput (Chapter 2), minimizing costs related to transmission power and delay (Chapter 3), and minimizing packet transmis- sion delay subject to maximum and average transmission power limits (Chapter 4). While designing these scheduling schemes, the random traffic and channel conditions are also taken into consideration. Our results show that the proposed opportunis- tic scheduling schemes can indeed take good advantage of the broadcast feature at the relay nodes and achieve much higher throughput and, in some scenarios, provide close-to-optimum QoS performance. The second part (Chapter 5) of the thesis deals with the issue of efficient resource management in multicast communications, where we study channel sharing and power allocations for multicast device-to-divice (D2D) communication groups underlaying a cellular network. In such a scenario, D2D multicasting together with the mutual inter- ference between cellular and D2D communications, makes the interference conditions and power allocations a very complicated issue. Different approaches are proposed that allow each D2D group to share the cellular channels and allocate transmission power to each D2D and cellular transmitter, so that the sum throughput of D2D and cellular users is maximized. Our results indicate that it is possible to achieve close-to-optimum throughput performance in such a network. / Dissertation / Doctor of Philosophy (PhD)

In accommodating the growing global demand for wireless, Multi-User Multiple-Input and Multiple-Output (MU-MIMO) systems have been identified as the key technology. In such systems, a transmitting basestation serves several users simultaneously, increasing the network capacity. However, sharing the same time-frequency physical resources can cause interference for the simultaneously scheduled users if not moderated properly. One way to mitigate this interference is by directing radio power through the radio channel in specific directions, a method which is called beamforming. Following the successful implementation of the AlphaZero algorithm in another radio resource management technique, scheduling, this thesis explores the potential of using a similar search-based method for the beamforming problem, striving towards the ultimate objective of making decisions for scheduling and beamforming jointly. However, as AlphaZero only supports discrete action spaces and the action space of the beamforming problem is continuous, a modification of the algorithm is required. The proposed course of action is to extend AlphaZero into Sampled AlphaZero, using sample-based policy improvement to create an algorithm that is both more scalable for large discrete action spaces and able to handle high dimensional continuous action spaces. To evaluate the performance of the models, test environments were simulated and solved using increasingly larger so-called codebooks, containing predefined beamforming solutions. The results of the Sampled AlphaZero model demonstrated promising performance even for very large codebook sizes, indicating the model's suitability for addressing the beamforming problem in a non-codebook-based context. Furthermore, this thesis explores how states in the search can be represented and preprocessed for the neural network to learn efficiently, demonstrating clear benefits of using a singular value decomposition-based state preprocessing over raw states as input to the neural network.

In recent years, game theory has emerged as a promising approach to solving the power control problem in wireless networks. This thesis extends the reach of game-theoretic analysis to embrace link adaptation, thereby constituting a generalization of the power control problem. A realistic and natural problem formulation is attempted, wherein transmitter power and a discrete-valued Adaptable Link Parameter (ALP), e.g. code rate, constitute the action set of a player in this game. The dual goals of maximizing throughput and minimizing power consumption are reflected in the utility function selection, which uses the accurate sigmoid model for approximating throughput. The discrete action space makes it difficult to verify the existence of a Nash Equilibrium (NE) in this game using standard techniques. To circumvent this limitation, a heuristic algorithm is proposed. This algorithm is analytically shown to always converge to a NE. The subsequent results probe its validity and sensitivity. Favorable comparisons are drawn between these game-theoretic results and those arising from parallel systems techniques. A linear programming system optimization that exploits properties of the dominant eigenvalue of the system gain matrix is also presented in a comparative context. / Master of Science

A thesis submitted in fulfilment of the requirements for the degree of Master of Arts, in ICT Policy and Regulation, University of the Witwatersrand, 2016 / There is a common opinion among researchers and experts that efficient management of radio spectrum plays a vital role in ensuring universal access to telecommunications services. The objective of this study was to identify radio spectrum reforms and their associated effects on market liberalisation. It was postulated that appropriate radio spectrum reforms would be catalysts for market liberalisation. The evolution of command-and-control approaches in relation to market-based approaches was assessed. The research involved literature critique, review of policies as that relates to history of radio spectrum management in South Africa and across the world, and radio spectrum regulations analysis in South Africa. Interviews of radio spectrum industry experts and documents study of the evolution of telecommunications regulatory environment with respect to radio spectrum management and market liberalisation were also used as main sources of research. The purpose of the literature critique, review of policies, regulations and documents was to identify hints of radio spectrum reforms and measure qualitatively the extent of market liberalisation. While interviews of radio spectrum industry experts were used to ascertain industry response to strides made as far as radio spectrum and market liberalisation in South Africa. It was observed that initially, in most parts of the world and in South Africa, market liberalisation progressed quickly despite appreciable correlation with radio spectrum reforms. Early radio spectrum reforms, such as the establishment of an independent regulator of the industry and radio spectrum, had contributed to some level of market liberalisation with creation of oligopolistic telecommunication market, and had increased to radio spectrum by Vodacom, MTN and Cell C having access to both 900 MHz and 1800 MHz bands. However, perpetual practise of command-and-control, an efficient radio spectrum management encouraged hoarding. The literature review and interview provided seven main contributions of reforms in the form of strides. These strides formed the basis for the research framework: 1) establishment of an independent regulator of the industry and radio spectrum, 2) increased access to radio spectrum, 3) service and technology neutrality on radio spectrum, 4) essential facilities to enable sharing, 5) market-based approaches radio spectrum pricing: AIP, 6) service-based competition versus infrastructure-based competition, and 7) non-rival, non-exclusive usage of radio spectrum. The conclusion is that increasing access to radio spectrum and the independent regulator were not primary determinants of market liberalisation. An analytic framework has been used to show that market liberalisation reached a plateau phase, with a few incumbents becoming dominant and creating an oligopolistic market structure. It is at this point that further market liberalisation could be stimulated by additional radio spectrum reforms. The command-and-control approach remains the main bottleneck source for access and efficiency in radio spectrum management, which encourages rival and exclusive use of radio spectrum. It has been observed that market-based radio spectrum reforms have also entrenched rivalry and exclusivity in the use of radio spectrum. Radio spectrum reforms that encourage non-rivalry and non-exclusivity, such as open-access to radio spectrum, are highly recommended in this research. / GR2016

Thesis (MSc.) -- University of Limpopo, 2018 / A Major concern in the past years was the traditional static spectrum allocation which gave rise to spectrum underutilization and scarcity in wireless networks. In an attempt to solve this problem, cognitive radios technology was proposed and this allows a spectrum to be accessed dynamically by Cognitive radio users or secondary users (SUs). Dynamic access can efficiently be achieved by making necessary adjustment to some MAC layer functionalities such as sensing and channel allocation. MAC protocols play a central role in scheduling sensing periods and channel allocation which ensure that the interference is reduced to a tolerable level. In order to improve the accuracy of sensing algorithm, necessary adjustments should be made at MAC layer. Sensing delays and errors are major challenges in the design of a more accurate spectrum sensing algorithm or MAC protocol. Proposed in this study, is a scheme (EXGPCSA) which incorporate sensing at the MAC layer and physical layer. Energy detector was used to detect the presence of primary users (SU). A choice of how long and how often to sense the spectrum was addressed at the MAC layer. The focal point of this study was on minimizing delays in finding available channels for transmission. EXGPCSA used channel grouping technique to reduce delays. Channels were divided into two groups and arranged in descending order of their idling probabilities. Channels with higher probabilities were selected for sensing. Three network scenarios were considered wherein a group of SUs participated in sensing and sharing their spectral observations. EXGPCSA was designed such that only SUs with higher SNR were allowed to share their observations with other neighbouring SUs. This rule greatly minimized errors in sensing. The efficiency of EXGPCSA was evaluated by comparing it to another scheme called generalized predictive CSA. A statistical t-test was used to test if there is significant difference between EXGPCSA and generalized predictive CSA in terms of average throughput. A test has shown that EXGPCSA significantly performed better than generalized predictive CSA. Both schemes were simulated using MATLAB R2015a in three different network scenarios.

The traditional way of increasing capacity in a wirelesscommunication system has been cell splitting and fixedchannel-allocation based on prediction tools. However, theplanning complexity increases rapidly with the number of cellsand the method is not suitable for the large temporal andspatial traffic variations expected in the future. A lot ofresearch has therefore been performed regarding adaptivechannel allocation, where a channel can be used anywhere aslong as the signal-to-interference ratio (SIR) is acceptable. Acommon opinion is that these solutions must be decentralizedsince a centralized one would be overly complex. In this thesis, we study the locally centralizedbunch conceptfor radio resource management (RRM) in aManhattan environment and show that it can give a very highcapacity both for outdoor users and for indoor users covered byoutdoor base stations. We show how measurement limitations anderrors affect the performance and wepropose methods to handlethese problems, e.g. averaging of measured values, robustchannel selection algorithms, and increased SIR margins. Wealso study the computational and signaling complexities andshow that they can be reduced by splitting large bunches, usingsparse matrix calculations, and by using a simplified admissionalgorithm. However, a reduction of the complexity often means areduction of the system capacity. The measurements needed for RRM can also be used to find amobile terminal's geographical position. We propose and studysome simple yet accurate methods for this purpose. We alsostudy if position information can enhance RRM as is oftensuggested in the literature. In the studied scenario, thisinformation seems to be of limited use. One possible use is toestimate the mobile user's speed, to assist handover decisions.Another use is to find the location of user hotspots in anarea, which is beneficial for system planning. Our results show that the bunch concept is a promisingcandidate for radio resource management in future wirelesssystems. We believe that the complexity is manageable and themain price we have to pay for high capacity is frequentreallocation of connections. <b>Keywords:</b>bunch concept, radio resource management,network-assisted resource management, base station selection,dynamic channel allocation, DCA, channel selection,least-interfered, interference avoidance, interferenceaveraging, handover, power control, path-loss measurements,signal strength, link-gain matrix, TD-CDMA, UTRA TDD, Manhattanscenario, microcells, mobile positioning, position accuracy,trilateration, triangulation, speed estimation

We develop efficient coordination techniques that support inelastic traffic in large-scale distributed dynamic spectrum access DSA networks. By means of any learning algorithm, the proposed techniques enable DSA users to locate and exploit spectrum opportunities effectively, thereby increasing their achieved throughput (or "rewards" to be more general). Basically, learning algorithms allow DSA users to learn by interacting with the environment, and use their acquired knowledge to select the proper actions that maximize their own objectives, thereby "hopefully" maximizing their long-term cumulative received reward/throughput. However, when DSA users' objectives are not carefully coordinated, learning algorithms can lead to poor overall system performance, resulting in lesser per-user average achieved rewards. In this thesis, we derive efficient objective functions that DSA users an aim to maximize, and that by doing so, users' collective behavior also leads to good overall system performance, thus maximizing each user's long-term cumulative received rewards. We show that the proposed techniques are: (i) efficient by enabling users to achieve high rewards, (ii) scalable by performing well in systems with a small as well as a large number of users, (iii) learnable by allowing users to reach up high rewards very quickly, and (iv) distributive by being implementable in a decentralized manner. / Graduation date: 2013