Login | Register

Performance Optimization of Cloud Radio Access Networks


Performance Optimization of Cloud Radio Access Networks

Labana, Mohamed (2021) Performance Optimization of Cloud Radio Access Networks. Masters thesis, Concordia University.

[thumbnail of Labana_MASc_S2021.pdf]
Text (application/pdf)
Labana_MASc_S2021.pdf - Accepted Version
Available under License Spectrum Terms of Access.


The exponential growth of cellular data traffic over the years imposes a hard challenge on the next cellular generations. The cloud radio access network (CRAN) is an emerging cellular architecture that is expected to face that challenge effectively. The main difference between the CRAN architecture and the conventional cellular architecture is that the baseband units (BBUs) are aggregated at a centralized baseband unit pool, hence, enabling statistical multiplexing gains. However, to acquire the several advantages offered by the CRAN architecture, efficient optimization algorithms and transmission techniques should be implemented to enhance the network performance. Hence, in this thesis, we consider jointly optimizing user association, resource allocation and power allocation in a two tier heterogeneous cloud radio access network (H-CRAN). Our objective is to utilize all the network resources in the most efficient way to maximize the network average throughput, while keeping some constraints such as the quality of service (QoS), interference protection to the devices associated with the Macro remote radio head (MRRH), and fronthaul capacity. In our system, we propose using coordinated multi-point (CoMP) transmissions to utilize any excess resources to maximize the network performance, in contrast to the literature, in which CoMP is usually used only to support edge users. We divide our joint problem into three sub-problems: user association, radio resource allocation, and power allocation. We propose matching game based low complexity algorithms to tackle the first two sub-problems. For the power allocation sub-problem, we propose a novel technique to convexify the non-convex original problem to obtain the optimal solution. Given the conducted simulations, our proposed algorithms proved to enhance the network average weighted sum rate significantly, compared to the state of the art algorithms in the literature.
The high computational complexity of the optimization techniques currently proposed in the literature prevents from totally reaping the benefits of the CRAN architecture. Learning based techniques are expected to replace the conventional optimization techniques due to their high performance and very low online computational complexity. In this thesis, we propose tackling the power allocation in CRAN via an unsupervised deep learning based approach. Different from the previous works, user association is considered in our optimization problem to reflect a real cellular scenario. Additionally, we propose a novel scheme that can enhance the deep learning based power allocation approaches, significantly. We provide intensive analysis to discuss the trade-offs faced when employing our deep learning based approach for power allocation. Simulation results prove that the proposed technique can obtain a very close to optimal performance with negligible computational complexity.

Divisions:Concordia University > Gina Cody School of Engineering and Computer Science > Electrical and Computer Engineering
Item Type:Thesis (Masters)
Authors:Labana, Mohamed
Institution:Concordia University
Degree Name:M.A. Sc.
Program:Electrical and Computer Engineering
Date:2 March 2021
Thesis Supervisor(s):Hamouda, Walaa
Keywords:research-creation Cloud radio access networks Deep learning Neural networks Cellular Networks 5G
ID Code:988227
Deposited By: Mohamed Labana
Deposited On:29 Jun 2021 22:28
Last Modified:29 Jun 2021 22:28
All items in Spectrum are protected by copyright, with all rights reserved. The use of items is governed by Spectrum's terms of access.

Repository Staff Only: item control page

Downloads per month over past year

Research related to the current document (at the CORE website)
- Research related to the current document (at the CORE website)
Back to top Back to top