Handover Reduction in 5G High-Speed Network Using ML-Assisted User-Centric Channel Allocation

In this paper, we propose a novel user-centric channel allocation scheme for high-speed terrestrial users of the Fifth Generation (5G) network in the millimetre-Wave (mm-Wave) band small-cells named Vehicular Frequency Reuse (VFR) scheme. To adapt the VFR scheme with the 5G network, we develop a new mobility management function that improves the 5G's performance for high-speed road users such as Connected Autonomous Vehicles (CAV) in small-cells by reducing the number of handovers (HOs) in the Vehicle-to-Network (V2N) service. The VFR scheme significantly reduces the users' HO rate, control plane signalling in air interface, and improves link reliability and channel reuse ratio. A metric called Distance-Threshold (DT) is defined to determine the frequency reuse ratio for the 5G network with the VFR scheme. We also propose a new cell reselection procedure for high-speed users in RRC_Connected (Radio Resource Control) state that are using the VFR scheme and managed by our mobility management function. The proposed cell reselection procedure is defined for inter-gNB-DU (gNodeB-Distributed Unit) and intra-gNB-DU mobility. This procedure reduces traffic load on the UE's (User Equipment) air interface, lowers processing and signalling load for network nodes, and assists for seamless mobility management for high-speed users. These all help and facilitate the path towards the targeted zero millisecond mobility interruption time (MIT) for 5G-NR (New Radio) users. Moreover, the proposed scheme, function, and procedure are compatible with the existing 5G structure and user equipment and can be easily added to the network by only software patches. The proposed mobility management function separates low-speed and high-speed users to serve them accordingly with different sets of channels. To separate high-speed and low-speed users, we propose a simple scalar metric defined as a Velocity-Threshold (VT). The VT value is adaptively calculated by a K-Means approach which is a well known unsupervised Machine Learning (ML) algorithm, according to the road condition inferred from reported velocities. Finally, we evaluate the proposed VFR scheme and compare it with the traditional cell-centric channel allocation scheme. Computer simulations show that the proposed VFR scheme can reduce the number of HOs (HO rate) for users by over 99% compared with the traditional scheme.