Poster: A Reinforcement Learning Approach for Base Station On/Off Switching in Heterogeneous M-MIMO Networks

The introduction of large antenna arrays facilitating massive multiple-input-multiple output (M-MIMO), and the addition of a tier of pico and femto base stations (BS), implementing an heterogeneous network, provides means to improve network throughput and capacity in 5G networks. However, the addition of antennas and BSs implies additional hardware and is associated with higher energy consumption. Improving the energy efficiency (EE) while reducing the power consumption of such heterogeneous M-MIMO dense networks can be performed by switching off base stations that have few users to serve and redistribute those users among the active neighboring base stations. One promising solution to intelligently map user spatial distribution to the optimal set of active BSs is by utilizing radio service maps (RSM). In this paper we propose a novel approach that effectively switches off base stations by combining reinforcement learning with RSM data. The proposed approach is evaluated through computer simulations using a 3D ray tracing model. The simulation results show the benefits of RSM and machine learning use for the improvements in EE of the considered heterogeneous M-MIMO networks.