Using Machine Learning to Locate Gateways in the Wireless Backhaul of 5G Ultra-Dense Networks
| Autor: | John W. Chinneck, Mital Raithatha, Roshdy H. M. Hafez, Aizaz U. Chaudhry |
|---|---|
| Rok vydání: | 2020 |
| Předmět: |
business.industry
Computer science ComputerSystemsOrganization_COMPUTER-COMMUNICATIONNETWORKS Monte Carlo method 020206 networking & telecommunications Topology (electrical circuits) 0102 computer and information sciences 02 engineering and technology Machine learning computer.software_genre 01 natural sciences 010201 computation theory & mathematics Logic gate Default gateway Computer Science::Networking and Internet Architecture 0202 electrical engineering electronic engineering information engineering Wireless Artificial intelligence business Cluster analysis Wireless sensor network computer 5G |
| Zdroj: | ISNCC |
| DOI: | 10.1109/isncc49221.2020.9297282 |
| Popis: | A distributed wireless backhaul has emerged as an attractive solution for forwarding traffic to the core in 5G Ultra-Dense Networks (UDNs). It consists of a large number of small cells and a few of these cells, referred to as gateways, are linked to the core by high capacity fiber optic links. Each small cell is associated to one gateway and forwards its traffic to it directly or through multiple hops. The backhaul network capacity increases by decreasing the average number of hops. In this paper, we consider two machine learning-based clustering algorithms, namely, k-means and k-medoids, to find gateway locations that minimize the average number of hops. We compare their performance with a baseline approach at different small cell densities through extensive Monte Carlo simulations in terms of average number of hops. The results indicate that both clustering algorithms significantly outperform the baseline approach and k-medoids performs equal to or better than k-means. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|