5G mmWave Multiple Active Protocol Stack Optimized Deep Learning Handover

To maintain seamless real-time connectivity for 5G applications, various handover (HO) schemes that can support a near-zero mobility interruption time (MIT) have been proposed. Due to the high frequency of mmWave signals, line-of-sight (LOS) is required for communication, because significant degradation and HO failure (HOF) occurs in non-LOS (NLOS) situations. In this letter, a novel deep learning based multiple active protocol stack (MAPS) optimized HO scheme is proposed to reduce the outage rate during the HO process and decrease the MIT. The performance evaluation results show that a much lower MIT and outage rate can be obtained when using the proposed MAPS HO scheme, compared to other HO techniques.