Connectivity-aware UAV mobility in cellular networks: DRL path planning and predictive handover

Abstract

The seamless integration of unmanned aerial vehicles (UAVs) into cellular networks promises transformative advances in logistics and emergency response, yet is fundamentally challenged by volatile airspace connectivity and inefficient handover mechanisms. To address these issues, we present a synergistic framework that unifies three-dimensional deep reinforcement learning (DRL)-based path planning with a predictive Reference Signal Received Power (RSRP)-aware handover algorithm. This framework jointly optimizes global path selection—using a multi-step dueling double deep Q-network (D3QN) to navigate challenging connectivity—and local handover management—employing a foresighted, proactive strategy to enhance handover efficiency. Such anticipatory decision-making ensures robust UAV communication under dynamic channel conditions. Unlike traditional separate optimization, it couples path planning with handover via future trajectory insights, enabling proactive adjustments. This integration yields superior performance, balancing throughput, outage risk, and handover overhead more effectively than isolated methods. Simulation results demonstrate substantial improvements over baseline methods in throughput, outage probability, and handover frequency across diverse flight scenarios. This work highlights that an integrated, anticipatory approach is crucial for developing scalable and resilient UAV communications in next-generation wireless networks.