Game Theory Based Resource Allocation Algorithm For D2D Communications In LTE-A Systems

With the development of mobile communications, the requirements on data rates are increasing sharply. The scarcity of spectrums is also becoming more and more serious. In order to get higher total capacity, an advanced technology must be put forward. Under this condition, Device-to-device (D2D) communication has been proposed and received increasing attentions as a promising component to improve spectrum efficiency. Unlike the infrastructure based cellular network, D2D users (user equipments or mobile terminals) do not communicate via the central coordinator (base station, NodeB or evolved NodeB) but operate as an underlay and communicate directly with each other or more hops. Excluding the unnecessary core network involvement, D2D communication is an appealing concept which can provide several advantages such as low power consumption, low cost, high bit rates and efficient frequency. However, these advantages hinge on the intelligent resource allocation approaches. This is because sharing the frequency between the D2D and cellular users makes the intra-cell and inter-cell interference inevitable. For example, when sharing the uplink resources the eNB and D2D receivers suffer the interference from D2D transmitters and cellular users, respectively. In the case of sharing the downlink resources, the cellular users and D2D receivers are exposed to the interference from D2D transmitters and the eNB. If the interference is not avoided effectively, the performance of D2D communications will be decreased, so that it leads to inefficient outcome of the whole system.To decrease the interference and increase the whole cell throughput, this thesis proposed two solutions based on the game theory for D2D communication in Long Term Evolution-Advanced (LTE-A) system. These two approaches are based on the scenario in which the D2D users share the uplink frequency with the cellular users and the cellular users employ the orthogonal resource. For the scenario where one D2D pair shares the same resource with one cellular user, a Stackelberg game based resource allocation scheme is proposed. In this game, the eNB and the D2D user make up a seller-to-buyer pair to improve the whole throughput, where the eNB is viewed as the seller and the D2D user is the buyer. The seller acts first, then the buyer observes the seller’s behavior and decides its strategy. In detail, the eNB charges some fees to the D2D user to increase its performance of every link, whereas the D2D user improves its rates by paying some fees for reusing the rasource. Finally, the game will lead to an equilibrium, making the cellular link and D2D link communicate reliably.For the scenario where one D2D pair can share the same resource with multiple cellular users and multiple D2D pairs can meanwhile employ same resources, an overlapping coalition formation game based resource allocation approach is proposed to improve the frequency efficiency and the sum throughput, which takes into account the Quality of Service (QoS) requirements of each user. Different from existing work which focused on the scenario that one cellular user shares spectrums with one D2D pair and thus disjoint groups are formed to use spectrums, in this thesis, we proposed a coalition formulation game with overlapping coalitions. In this game, we model the resource sharing problem as a transferable coalition formation game and the D2D users can make a decision to join or leave a coalition according to the merge-and-split rule. Consequently, not only one D2D pair can employ the same resource with multiple CUEs, but also the channel of one cellular user can be allocated to multiple D2D pairs. From the proposed approach, we can find the stable coalitions as the result of the game. And a discrete-time Markov chain-based analysis is presented to obtain the stable coalitions. In particular, we considered the practical case and studied the effect of imperfect channel state information (CSI) on the stable coalition formation game. The complexity of employing this scheme is evaluated at the end of the thesis. And simulation results prove that satisfying performance can be achieved by using the proposed mechanisms.The contributions are summarized as follows.1. We studied game theory based resource allocation algorithms for D2D communications. In the Stackelberg game based resource allocation scheme, we group the eNB and the D2D user into a seller-to-buyer pair to improve the whole throughput.2. We also proposed the policies to resolve the conflict when multiple D2D users select the resources.3. In particular, unlike the traditional non-overlapping coalition formation game where one D2D pair joins only one coalition and different resources are allocated orthogonally among these coalitions, we proposed a solution based on an overlapping coalition formation game in which the D2D pair can access into different coalitions to improve the frequency efficiency and the sum throughput which takes into account QoS requirements of each user. This scheme is particularly suitable for a general case where there are multiple potential D2D pairs underlaying cellular networks. We also model the uplink resource sharing problem as a transferable coalition formation game and the D2D pairs make their decisions autonomously to join or leave a coalition according to the merge-and-split rule. It allows the users to flexibly self-organize themselves into a stable partition, as well as adapt to the dynamic environments. The proposed game-theoretic framework accounts for the inherent tradeoff between the sum rate and the mutual interference. Particularly, we take advantage of a discrete-time Markov chain to analyze the stability of the coalition formation.4. Since it is difficult to have complete channel state information (CSI) at the eNB and UEs, we studied the effect of imperfect CSI on the stable coalition formation game. The complexity of employing this scheme is evaluated at the end of the thesis. Extensive simulation results are provided to demonstrate the effectiveness of our proposed game model and algorithm.