Skku| autonomous handover decision of UAV Based on deep reinforcement learning

【 title 】Proactive Handover Decision for UAVs with Deep Reinforcement Learning

【 The author team 】Younghoon Jang, Syed M. Raza, Moonseong Kim, Hyunseung Choo

【 Date of publication 】2022.2.5

【 Thesis link 】https://www.mdpi.com/1424-8220/22/3/1200/htm

【 Recommended reasons 】 Unmanned aerial vehicle (UAV) The application of is in monitoring 、 Areas such as logistics and entertainment are growing rapidly , Continuous connectivity to the cellular network is required to ensure seamless operation . However , At present, the handover strategy in cellular network is mainly designed for ground users , Due to the frequent fluctuation of air signal strength , Not applicable to drones . In this paper, a new handoff decision scheme is proposed , The scheme deploys deep reinforcement learning (DRL), To prevent unnecessary switching , While maintaining stable connectivity . The proposed DRL The framework takes the UAV state as the input of the near end strategy optimization algorithm , The received signal strength index is developed based on the reward function (RSSI), It is used for online learning UAV switching decision . The proposed scheme is in 3D Simulation UAV Evaluate in a mobile environment , With greed and based on Q Study of the UAV Switching decision scheme , It reduces by up to 76% and 73% Unnecessary switching .