Machine Learning-Based MIMO Enabling Techniques for Energy Optimization in Cellular Networks

In this paper, we consider the problem of energy optimization in mobile networks by enabling the MIMO feature only when necessary. Enabling MIMO features at the base station increases energy consumption unnecessarily under many operating conditions. In this study, we employ machine learning-based approaches to decide on whether a SISO scheme can achieve the required Quality of Experience (QoE). If SISO can satisfy the target QoE, the base-station can decide to switch the MIMO feature off which can result in considerable energy savings. We consider two different machine learning approaches, namely, multi-layer perceptron (MLP) and recurrent neural networks (RNNs), to learn the SISO features from realistic mobile network data. The trained models are tested against the data obtained from MIMO cells in which the MIMO feature is disabled. Our results show the effectiveness of our proposed approach which presents a real-time, automated approach for MIMO enabling decisions.