| Summary: | The fundamental principle of multiuser diversity states that it is possible to achieve a significant received energy gain in a multi user wireless communication system if channel resources are divided between users in an adaptive, dynamic and intelligent fashion. When every channel resource block is assigned to the user with the best channel quality at any time or frequency bin, the system-wide sum rate achieved can be significantly increased. There is a risk, though, that users with a weaker channel quality might starve for resources, as this approach prioritizes the stronger users. Therefore, it is of utmost important that some type of fairness control mechanism is used at the same time. Most of the work published in the relevant literature assumes that a central controller dictates the resource allocation process. Such an approach induces various costs though; all users have to transmit their channel quality information to the central controller and then the allocation decisions have to be fed back to them. In an inherently fast-varying wireless channel this generates a lot of overheads and makes the whole allocation process less efficient. In an attempt to minimize this cost and to remove the need for the existence of a central controller, this PhD work investigates the use of game theoretic concepts to achieve distributed resource allocation. Coalition formation is a game theoretic concept that allows individual entities to cooper- ate and perform a task more efficiently. By allowing users to form disjoint coalitions in a distributed fashion, this work tackles the problem of cooperative distributed resource allocation. Using the Nash Bargaining Solution as a means for enabling intra-coalition cooperation, individual users are able to exchange resource blocks and harvest the bene- fits of multiuser diversity, in a fashion that maintains a level of fairness similar to that of Proportional Fairness. For the purposes of this work, three different schedulers for resource allocation were de- veloped and are presented in this thesis. Two of them are based on the aforementioned game theoretic concepts, while the third is a hybrid between the Greedy scheduler and the game theoretic approach. Simulation results are presented and comparison to the widely adopted Proportional Fair scheduler is made. Additionally, the trade-offs between sum rate performance, fairness, complexity and overheads are investigated and the protocol used for the allocation process is presented.
|
|---|
