Path Planning for Cellular-connected UAV Using Parabolic Equation-based Radio Wave Propagation Models

The high flexibility and mobility of cellular-connected unmanned aerial vehicles (UAVs) make them an important component of future wireless communication systems. However, planning their flight path is a challenging task, due to weak coverage regions in the sky and limited flight time. In this paper, we propose a novel approach that combines site-specific radio wave propagation models and deep reinforcement learning techniques to optimize the flight path for UAVs. Our approach aims to minimize the flying time while ensuring reliable communication connectivity with base stations. To that end, a parabolic equation method is employed to model radio wave propagation, while deep Q networks are applied to determine the optimal path. The proposed approach takes into account physical environmental factors and eliminates the need for time-consuming and labor-intensive measurement campaigns. We present numerical results for various random initial flying points, demonstrating the effectiveness of our approach in path planning for cellular-connected UAVs.