Distributed Algorithms for Adaptive Sleep Modes in 5G Networks

 

5G mobile networks will be densely deployed to cope with heavily varying traffic demand. This enables for strategies of energy savings when the utilization is low. Due to hardware properties base stations may sleep in different modes. The deeper the mode the more energy may be saved, but at the cost of longer periods of going to sleep and waking up again. The decisions which sleep mode to choose may me made by means of distributed algorithms from reinforcement learning, e.g., Q-Learning.
Within this project students will learn basics on 5G mobile networks and on reinforcement learning applying these basics on that problem.
A simulation environment with GUI will be developed, e.g., in Matlab, such that results on own examples may be evaluated.

Literature

• A Distributed Q-Learning Approach for Adaptive Sleep Modes in 5G Networks, Ali El-Amine, Mauricio Iturralde, Hussein Al Haj Hassany and Loutfi Nuaymi
• Reinforcement Learning - An Introduction, Sutton, Richard, Barto Andrew, MIT Press