1. UKM Learning and Research Repository
2. PHD Thesis / Tesis PHD
3. Faculty of Engineering and Built Environment / Fakulti Kejuruteraan dan Alam Bina

Title:	Traffic load balancing techniques based on Q-learning controlled cell range extension for LTE-Advanced Heterogeneous Networks
Authors:	Sameh Fathi Sadeq Musleh (P55891)
Supervisor:	Mahamod Ismail, Prof. Dr.
Keywords:	Long-Term Evolution (Telecommunications)
Issue Date:	May-2018
Description:	One of the main features in Long Term Evolution-Advanced (LTE-A) technology is the Heterogeneous Network (HetNet), which utilizes small cells to enhance the system capacity and coverage. The intensive deployment of small cells such as pico- and femtocells to complement macro-cells resulted in unbalanced distribution of traffic-load among cells. Cell Range Extension (CRE) is one of the Cell-Selection techniques in HetNets which balance traffic load by adjusting the distribution of end-users among macro-cells and small-cells. However, conventional CRE uses manually pre-defined power offset, and each type of small cells needs different offset value depending on end-users hotspot size, thus traffic-load is not optimally distributed among cells. This will result in higher rates of rejected end-user calls and higher interference level particularly at the cell edge. Moreover, CRE is not self-tuned to support dense and self-organized HetNet. The main goal of this thesis is to propose traffic load balancing techniques based on Q-learning controlled CRE for HetNet. Firstly, two algorithms are proposed named as Load Balancing based on Q-Learning of end-user SINR (LBQS) and Load Balancing based on Q-Learning of Cell-throughput (LBQT). Both algorithms control the reference signal power of each small-cell that underlays highly loaded macro-cell while monitoring any possible degradation in the performance metrics of both cells, and reacts to troubleshoot any performance degradation in real time during each optimization epoch. Secondly, an improved version of the algorithms named, Load balancing based on Mamdani Fuzzy QLearning (LBMFQ) and Load Balancing based on Sugeno Fuzzy Q-Learning (LBSFQ) has been proposed. The Fuzzy Q-learning based Controller interacts continuously with large number of states and actions and feeds the network with new parameter settings. The proposed techniques have been validated via simulation and benchmarked against the conventional CRE in term of cell throughput, Signal-to-Interference-and-Noise Ratio (SINR) and spectral efficiency. All algorithms showed an improvement of 15%, 24% and 31% in cell throughput, spectral efficiency and end-user SINR respectively. Both LBMFQ and LBSFQ are further improved by 10% at least in comparison with LBQS and LBQT algorithms. The proposed load balancing techniques shall be improved to accommodate the ultra-dense small cells traffic in the future Fifth Generation networks.,Ph.D.
Pages:	126
Call Number:	TK5103.48325 847 2018 3 tesis
Publisher:	UKM, Bangi
Appears in Collections:	Faculty of Engineering and Built Environment / Fakulti Kejuruteraan dan Alam Bina

Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.