无人机辅助边缘计算环境下的路径策略研究

Research on Path Planning in Unmanned Aerial Vehicle-Assisted Edge Computing Environments

摘要: 无人机(Unmanned Aerial Vehicles, UAVs)越来越多地被用作移动边缘计算(Mobile Edge Computing, MEC)中的移动服务器，因为它们能够为计算能力有限的终端用户提供低延迟、高可靠性和强大的计算服务。在具有挑战性的环境当中，现阶段进行了广泛的研究，提出了各种方法来增强无人机在这种条件下的适应性，并改善提供给地面终端用户的服务质量(Quality of Service, QoS)。本文在现有研究的基础上引入了更复杂的环境因素，以确保更接近真实世界的复杂性和多变性。此外，本文在研究无人机在复杂环境中的飞行路径过程中，通过考虑静态和移动障碍物引起的风险因素、终端用户需求的动态变化以及与无人机飞行能耗因素的影响。为了实现这一目标，本文将这些因素的影响纳入奖励矩阵中，并提出了相应的算法。本研究的目标是基于这些条件下，在更复杂的环境中验证目标算法的有效性和可靠性，并证明引入一个真实世界的环境的必要性。实验结果表明，所提目标算法在复杂环境中表现出鲁棒性和可靠性，优于其他基准算法，并强调模拟一个与真实世界密切相似的环境的重要性。

Abstract: Unmanned Aerial Vehicles (UAVs) are increasingly being used as mobile servers in mobile edge computing (MEC) due to their ability to provide low latency, high reliability, and robust computing services to terminal users with limited computational capabilities. There has been extensive re-search conducted on this technology in challenging environments, with various methods proposed to enhance the adaptability of UAVs under such conditions and improve the Quality of Service (QoS) provided to ground terminal users. This paper builds upon existing research by introducing more complex environmental factors to ensure a closer approximation to the real-world complexity and variability. Furthermore, this paper investigates the flight paths of unmanned aerial vehicles (UAVs) in complex environments by considering the influence of risk factors caused by static and moving obstacles, dynamic variations in end-user demands, and energy consumption factors asso-ciated with UAV flight. To achieve this objective, the impact of these factors is incorporated into the reward matrix to develop the proposed algorithm. The objective of this research is to validate the effectiveness and reliability of the proposed algorithm in a more complex environment, based on these conditions, and to demonstrate the necessity of introducing a truly realistic environment. Ex-perimental results indicate that the proposed algorithm exhibits robustness and reliability in com-plex environments, outperforming other benchmark algorithms, and highlighting the significance of simulating an environment that closely resembles the real-world.