A Handover Assistance Algorithm Based on SDN and Machine Learning

Abstract

In the 5th generation mobile network (5G), microcells are densely deployed for spatial multiplexing, working in concert with traditional macrocells. 5G network uses not only the sub-6GHz band, but also the millimeter wave (mm-Wave) band, while other radio access technology (RAT) such as long-term evolution (LTE) and LTE advanced (LTE-A), collectively referred to as LTE from now on, will continue to use the sub-6GHz band. Therefore, there is a measurement gap before performing handover (HO) from LTE to 5G. A measurement gap is the duration for which user equipment (UE) suspends communication with the serving base station (BS) and then measures adjacent frequencies or other adjacent RATs. The throughput reduction and handover failure (HOF) problems caused by measurement gaps need to be solved urgently. So, for BSs supporting cellular communication in sub-6 GHz and mm-Wave frequency bands, this paper proposes a HO assistance method based on software defined network (SDN) and extreme gradient boosting (XGBoost) algorithm. This method effectively avoids measurement gaps and reduces HOF by predicting the future BS selection of mobile UE through machine learning. The simulation results show that compared with the standard HO algorithm, the HO assistance strategy proposed in this paper can greatly improve the HO success rate.