Fully Distributed Approach for Energy Saving in Heterogeneous Networks

In this paper, we study the problem of energy consumption in Heterogeneous Networks (HetNets). With the billions of connected users, the use of HetNets is one of the promising solutions for future cellular technologies as they can handle high network capacity. The aim of this work is to study and improve the energy efficiency in HetNets. The underlying idea is to perform the energy saving process on a fully distributed manner without the need to know the global network state or to exchange messages. We first express the optimization problem where the optimal cell range expansion bias must be found to maximize the energy efficiency. Seen that the problem is NP- hard, we introduce a learning approach based on Multi-Armed Bandit (MAB) allowing the operating cells to learn, through multiple iterations, the best expansion bias to use.