UAVs Handover Decision using Deep Reinforcement Learning

Cellular networks provide the necessary connectivity to the Unmanned Aerial Vehicles (UAV), however, these net- works are primarily designed for ground users. The in place handover decision mechanism for ground users is inappropriate for UAV due to frequent fluctuations in signal strength. This paper proposes a Deep Reinforcement Learning (DRL) based UAV Handover Decision (UHD) scheme to determine when it is essential for UAV to execute the handover for maintaining stable connectivity. DRL framework uses Proximal Policy Optimization algorithm to dynamically learn the UHD in an emulated 3D UAV mobility environment to manage the handover decisions. Experimental results show that UHD reduces handovers up to 76% and 73% comparing to conventional and target methods, respectively, while maintaining signal strength for stable and reliable communication.