Abstract
In light of recent interests and activities by cellular networks operators to exploit the unlicensed bands to boost network capacity, never has the issue of fair coexistence and spectrum sharing been at the heart of most feasibility and performance studies of radio access technologies coexisting in unlicensed bands. This has been the case because overall system performance by incumbent radio access technologies is not expected to be compromised in the spirit of fairness. Therefore, fair coexistence or spectrum sharing has become a key performance metric in evaluating the performance of most solutions proposed to permit radio access technology coexistence in the unlicensed bands. Time- based fairness, the focus of this thesis, refers to a mechanism which can by adopted to evaluate fairness performance among coexisting radio access technologies in the time domain, however, limited studies have been conducted and practical implementation solutions are still out of reach. In this thesis, the objectives are to address existing gaps in accessing practical time-based fairness solutions. Firstly, a review of the state of the art on fairness issues, metrics and approaches are discussed, providing an overview of current approaches and solutions and identifies their shortcomings to practical implementation. Secondly, the estimation of number of nodes contending over the unlicensed spectrum, which is a requirement for many fairness oriented schemes proposed for radio access technologies coexisting in unlicensed bands, is addressed. A novel technology-neutral estimation method for node numbers is proposed. The transmission interval observed over the unlicensed channel represents a probabilistic distribution, which can be obtained via the uniform difference distribution. The characteristic features of the uniform difference distribution are exploited to aid estimation of node numbers in scenarios where nodes contend for the channel within the same contention window and under multiple contention windows. The benefit of the proposed method over existing methods is the level of accuracy and its ability to provide a tighter estimate to small increase in numbers of contenders on the channel. Thirdly, two approaches to achieving time-based fairness are proposed. The first being a deterministic approach and the second a probabilistic one. The deterministic approach aims to study the upper bound performance of time-based fairness utilizing the estimation method for node numbers proposed in this thesis. An optimal value is computed as a backoff value for each transmission cycle while keeping the transmission opportunity of all coexisting radio access technologies the same. The results show time-based fairness improves spectrum utilization and overall throughput performance. The probabilistic approach seeks for a practical and implementable solution to achieving time-based fairness based on the proven performance benefits shown from the deterministic approach. The transmission interval distribution obtained from observation of the channel activity during the node number estimation serves as the building block towards a practical solution. The distribution is continually monitored and mapping of the occurrence of transmission intervals to its probability density functions are performed. The mean of the distribution is then optimized to provide a balance between the channel access probabilities in order to achieve approximately equal transmission time by all nodes contending over the channel. Two parameters are adjusted to attain time-based fairness, which are the contention window sizes and transmission opportunity. Simulation results show that time-based fairness under the proposed scheme can improve spectrum utilization, reduce the disparity in throughput performance and guarantee fairness among coexisting radio access technologies.