Advanced Conditional Handover in 5G and Beyond using Q-Learning

The vision of the new cellular technologies is to provide seamless wireless connectivity to different market verticals with diverse quality of service requirements. Conditional Handover (CHO) is a Third Generation Partnership Project (3GPP) feature designed to enhance the mobility robustness and it is supposed to ensure seamless connection for end-users in mobility scenarios. Even though CHO provides handover configuration to the User Equipments (UEs) well in advance, sometimes UEs still face service interruptions caused by Radio Link Failure (RLF). One of the prime reason is that practically the base stations cannot provide the ideal handover parameters to the User Equipments which can work perfect in all the future conditions of the UE. In order to solve this problem, we propose a novel method based on Epsilon-Greedy Q-learning called Advanced Conditional Handover (ACHO) which allows the UE to compute optimal handover parameters based on prevailing conditions and previous handover results and apply for the CHO target cells. ACHO is simulated for different scenarios involving UE speed, slices and signal conditions. From the simulation results, there is significant improvement observed in quality of handover decisions made with ACHO.