Dynamic spectrum allocation using game theoretical approaches.

Abstract

The increasing demand for spectrum due to the growing wireless services causes spectrum shortage and thus motivates changes in the way the spectrum resources are managed. With the limited spectrum available, mobile users will have to compete for frequency channels to transmit data. Such problems can be formulated as games where the players involved interact with one another, each with the objective of maximizing its own utility function selfishly. This thesis aims to adopt game theory in devising efficient radio resource management (RRM) algorithms for dynamic spectrum allocation in wireless communication systems.

The RRM problem in this research is envisioned by using a two-tier framework. At the lower tier, the thesis considers games where players are the wireless devices, such as mobile users, access points, base stations, etc., which have data to transmit and need to access a common pool of the shared spectrum. For such systems, the underlying factor that may cause user performance degradation is co-channel interference (CCI). Consider a scenario with multiple transmitter-receiver pairs in an ad-hoc orthogonal frequency-division multiple access (OFDMA) network, where each pair of users needs to select OFDMA channels and a certain level of power to transmit in a distributed manner. In order to combat CCI, this problem is modeled as a strategic-form game with the utility function designed to optimize the signal-to-interference-and-noise (SINR) ratio. Iterative algorithms are devised to obtain the Nash equilibrium (NE), which is the stable operating point of the system.

The iterative best-response techniques are frequently used in game theory but they might not guarantee convergence, especially when pure-strategy NEs might not exist. However, they do exist for certain classes of games, such as potential games. By adopting an interference-minimizing utility function, a potential game for the OFDMA system under investigation can be formulated. The results are further extended to an infrastructure-based multi-cell OFDMA network where a player consists of a base station and multiple mobile users in the cell. As such, pure-strategy NEs exist and can always be reached by using the best/better-response dynamics. Other issues such as the fairness among the center-users and edge-users in cellular OFDMA systems or the optimality analysis of the algorithms will also be addressed in this research. These methods are able to satisfy the users' target SINR and system fairness in the numerical simulation studies.

At the upper tier, the thesis looks at a different scenario where the competition involves multiple licensed and unlicensed spectrum providers. The primary spectrum holders (PSHs) are the licensed spectrum users, e.g., operators in a geographical area, who are allowed to sell/lease portions of their available bandwidth to other unlicensed secondary service providers (SSPs) for monetary profits. Reallocating spectrum and encouraging sharing among the licensed and unlicensed service providers are new RRM methodologies proposed in future dynamic spectrum access systems, where micro-economic mechanisms could be employed to study the emerging spectrum market. This problem is viewed as an oligopolistic differential game between multiple PSHs, managed by a controller called the spectrum broker. By assuming the spectrum price is a dynamic continuous-time process governed by a differential equation, the proposed method seeks to find a Markov NE solution and determines the market equilibrium price. In addition, a discrete-time price adjustment which can be easily implemented at the spectrum broker is proposed. Extensive numerical studies are carried out to investigate the dynamic behaviors of the spectrum market.