Dynamic Construction and Adaptation of 3D Virtual Network Topology for UAV-Assisted Data Collection

Abstract

Unmanned aerial vehicles (UAV) assisted data collection from on ground devices and sensors is becoming more useful in many mission-critical applications. However, meeting the data collection requirements under dynamic channel conditions between the UAV and on ground devices relies on frequent information exchanges, which brings great challenges to the dynamic operation of the integrated UAV network due to its inherent complexity. To rapidly obtain a holistic view in assisting the UAV network operation, we first propose a three-dimensional (3D) virtual network topology which helps the UAV to make faster decisions by analyzing refined virtual indicators instead of measuring and processing related physical factors frequently in real time. To improve the efficiency of UAV data collection, dynamic adaptation of the 3D virtual network topology is achieved by a deep deterministic policy gradient (DDPG) based algorithm, where the UAV flying speed and direction, as well as the determination of the target group of on ground devices are optimized under the UAV energy constraint. Simulation results demonstrate that the proposed DDPG-based dynamic adaptation of the 3D virtual network topology can effectively improve the data collection efficiency compared with the benchmark solutions.