Open Collections

Machine learning-assisted CRAN design with hybrid RF/FSO and full-duplex self-backhauling

University of British Columbia

The ever increasing demand for higher data rates, lower latency communication, and a more reliable mobile network has led us toward the 5th generation (5G) of mobile networks. In 5G, resource allocation is one of the most challenging problems. Conventionally, model-driven methods, and analytical approaches have been used to allocate resources optimally. Despite accuracy, these methods often result in a non-convex optimization problem that is inherently challenging to handle and require proper convex approximation. To overcome such drawbacks, we need more efficient resource allocation techniques in the 5G mobile network. This research will study the downlink of a cloud radio access network. The cloud radio access network enables coordinated beamforming and better interference management in ultra-dense networks. This architecture's bottleneck is backhaul capacity restriction limiting the benefits that the cloud radio access network offers. We will use hybrid radio frequency and free-space optical links to address the backhaul capacity limitation. Also, to improve the throughput and increase the spectral efficiency of the radio-frequency links, we propose in-band full-duplex self-backhauling radio units. After formulating the mathematical model and solving it with analytical approaches, we will introduce a novel solution for the proposed scenario and show that it outperforms the state-of-the-art half-duplex backhaul technology provided enough self-interference cancellation under various weather conditions. We will derive a joint optimization problem to design the backhaul and access link precoders and quantizers subject to the fronthaul capacity, zero-forcing, and power constraints. We will show that this problem is non-convex and computationally intractable and approximate it with a semi-definite programming that can be effectively solved by alternating convex optimization. We also employed Compute Canada computational resources for solving mentioned semi-definite programming. The computational complexity of the proposed optimization approach motivates us to employ machine-learning-based optimization methods that recently received much recognition in academia and industry. We use supervised and unsupervised deep neural networks for learning the optimal resource allocation strategy and achieved 80% of the performance compared to the proposed analytical approach with only a fraction of computational cost. To meet all feasibility constraints of the problem, we also propose customized activation functions and post-processing steps.

Text

2020-12-02

Attribution-NonCommercial-NoDerivatives 4.0 International

http://hdl.handle.net/2429/76630

Electrical and Computer Engineering

University of British Columbia

UBCV

http://creativecommons.org/licenses/by-nc-nd/4.0/