چكيده لاتين
Ultra-dense networks with full-duplex communication (FD-UDNs) is a promising technology in cellular networks, providing high capacity and massive access. The high density of small cells in these networks enhances both the capacity and coverage of the network, while full-duplex communication has the potential to double the network capacity by enabling simultaneous transmission and reception in the same frequency channel. However, these networks suffer from high energy consumption and severe interferences, which can negatively affect the overall efficiency of the network. The conducted research proves that resource allocation and base station (BS) sleep has a significant contribution in managing interference and reducing energy consumption, especially in dense deployment of small cells. Therefore, it is necessary to utilize BS sleep strategy along with resource management techniques to effectively reduce interference and power consumption in FD-UDN, which has not been considered in previous works. To this end, this thesis takes into consideration an EE maximization through jointly optimizing resource management and BS sleeping, where the data rate requirements of both uplink and downlink users are met. The formulated optimization problem is a nonlinear, mixed integer non-convex programming problem which is difficult to solve. To tackle this problem first, a centralized solution based on classical optimization techniques has been presented. In this solution, at first the objective function is transformed into an equivalent parametric subtractive form by using the Dinkelbach method. Then, the problem is decoupled into two sub-problems: 1)problem of user association and resource allocation, which is transformed by using some appropriate transformations into a convex optimization form and then solved, and 2) problem of optimal BSs on/off switching, which is solved by using the dual Lagrangian method and the CCCP method. In the following, considering the computational and signaling overhead of this solution as well as the high dynmics of wireless networks, a solution based on reinforcement learning method is introduced. In second solution, also the problem is divided into two sub-problems: resource allocation and BS sleep management. In BS sleep management, a centralized and a distributed solution have been proposed. In distributd solution, each small base station (SBS) decides about its sleep status according to criteria such as EE, satisfaction rate (SR), full-duplexity rate (FDR), and free sub-channel rate (FSR). While in centralized solution, macro base station (MBS) decides about the sleep status of SBSs base on similar criteria. Then, by re-associating users to active base stations, each BS allocates transmission power and radio resources to its users. The simulation results demonestrates the effectiveness of the proposed algorithms in improving the energy efficiency, reducing the energy consumption and keeping fewer SBSs active, especially in high network loads.
