Deep Learning–Based Coverage and Capacity Optimization

This chapter presents two state‐of‐the‐art machine learning (ML)‐based techniques that tackle the coverage and capacity optimization (CCO) problem from each of the main aspects: configuring base‐station parameters to address current demand through a deep neural network architecture, where suitable configurations actions are taken on the basis of the inference from current network user geometry information; and enabling base‐station sleeping via a data‐driven approach by using deep reinforcement learning (RL), which leverages network traffic models to address the non‐stationarity in real‐world traffic. The chapter introduces a set of widely used ML techniques and provides an overview of their application to CCO problems in the wireless network domain. It then describes the used and achieved result of the deep RL approach in solving the problem of base‐station sleeping. The chapter also presents the application and evaluation of the multi‐agent deep neural network framework on the dynamic frequency reuse problem in mobile networks.