Dynamic Network Re-dimensioning via Accurate Prediction of 5G Throughput and Reinforcement Learning

The emerging 5G technology has tremendous potential in empowering a wide range of demanding applications. However, the 5G throughput can fluctuate wildly based on several factors like handoffs, user moving speed, user moving direction with respect to the radio base station, the terrain (e.g., obstacles between the user and the radio base station), the density of users and the workloads covered by the radio base station, the temporal patterns of usage (e.g., time of the day, days of the week, etc.), and more. These fluctuations in the throughput can sometimes drop below the throughput of 4G or sometimes to nearly zero (5G dead zones). It is of great importance to a 5G provider to better understand how any of these factors affect its end-users to provide better service. This project focuses on how to accurately model and predict the throughput of 5G network, and how to better manage the network’s resources to adapt to the changes in the patterns of usage.

Faculty Supervisor:

Ahmed El-Roby

Student:

Partner:

Ericsson Canada Inc (Quebec)

Discipline:

Computer science

Sector:

Information and cultural industries; Professional, scientific and technical services

University:

Carleton University

Program: