Practical Optimizing UAV Trajectory in Wireless Charging Networks: An Approximated Approach

Unmanned Aerial Vehicles (UAVs) can be easily deployed as auxiliary base stations due to their convenience and flexibility. However, limited battery capacity becomes a bottleneck. Promising wireless power transfer (WPT) technologies can provide a continuous power supply for UAVs. Many of the recent works treat the UAV battery capacity as a constraint, which hinders the assurance of continuous UAV operation. Furthermore, most studies employ intelligent path-planning algorithms that lack explicit performance guarantees. In this paper, we study the problem of Practical Optimizing UAV Trajectory in Wireless Charging Networks (POTWCN), which involves planning the trajectory of the wireless-powered UAV in the practical environment with obstacles by selecting candidate passing positions and determining the access order in the charging network. The goal is to maximize the benefit, i.e., balancing the total task completion time and the number of charging stations visited, so as to minimize path length and flight time, and ensure energy constraints with performance bound. To solve this problem, we first formalize the problem and prove its submodularity. Then, we propose the obstacle-aware weighted graph generation algorithm (OWGGA) to deal with the obstacles in the environment, which forms an obstacle-avoidance path using tangents and arcs between two hovering positions and the blocking obstacles. Next, we propose a dynamic charging station selection algorithm (ACSA), which maximizes the UAV’s energy utilization by limiting the number of charging stations that can be included. In the algorithm, we introduce the Christofides algorithm and use the path length calculated by OWGGA as the edge weights of the graph. Subsequently, considering the UAV’s energy constraints, we iteratively solve the UAV trajectory planning problem by adding the charging station with a maximized marginal benefit to the path. We prove that the proposed algorithm achieves an approximation ratio $1 - 1/e$ as well as the path length is at most $3\pi /4$ times the optimal solution. Simulation results show that our algorithm reduces the flight distance by 38.01% and the task completion time by 34.00% on average.