Data enabled Self-Organizing Network with Adaptive Antennas based on Proactive Prediction for Enabling 5G

Abstract

The increased energy consumption and operational expenditure (OPEX) that result from the heavy terminal load and network densification increase the cost of 5G networks. Using machine learning algorithms more potential information may be obtained from the user-related data that can play a pivotal role in automatic resource management enabling self-organizing networking (SON) in 5G networks. This paper proposes an energy efficiency (EE)-oriented self-optimization function based on proactive predictions in a multi-cell massive multiple-input multiple-output (MIMO) network. Our self-optimization function can automatically adjust the active antennas in each cell based on the predicted quantity of users in the next time interval to proactively adapt to the fluctuations in the traffic load. Furthermore, a two-step optimization scheme is proposed. The first-step independent optimization scheme employs the single-cell proactive prediction result to obtain the appropriate number of active antennas in each cell to maximize EE. The second-step joint optimization scheme employs multi-cell information related to the number of active antennas and users to further optimize the network. We use a real dataset from existing base stations (BSs) to test our prediction model and optimization schemes. The simulation results demonstrate that both schemes can yield a considerable performance improvement in EE compared with the static antennas adjustment scheme.