航空学报 > 2024, Vol. 45 Issue (8): 329054-329054-1   doi: 10.7527/S1000-6893.2023.29054

面向物联网数据收集的无人机自主路径规划

张薇(), 何若俊   

  1. 哈尔滨工程大学 信息与通信工程学院,哈尔滨 150001
  • 收稿日期:2023-05-29 修回日期:2023-07-12 接受日期:2023-11-06 出版日期:2024-04-25 发布日期:2023-11-16
  • 通讯作者: 张薇 E-mail:zhangwei@hrbeu.edu.cn
  • 基金资助:
    电子信息系统复杂电磁环境效应国家重点实验室资助课题(CEMEE2021K0101A)

Autonomous trajectory design for IoT data collection by UAV

Wei ZHANG(), Ruojun HE   

  1. College of Information Communication Engineering,Harbin Engineering University,Harbin 150001,China
  • Received:2023-05-29 Revised:2023-07-12 Accepted:2023-11-06 Online:2024-04-25 Published:2023-11-16
  • Contact: Wei ZHANG E-mail:zhangwei@hrbeu.edu.cn
  • Supported by:
    State Key Laboratory of Complex Electromagnetic Environment Effects on Electronics and Information System(CEMEE2021K0101A)

摘要:

路径规划旨在为无人机(UAV)提供安全可靠的行进路径,而实际环境的动态性极大地增加了路径规划的难度。本文针对物联网(IoT)节点处的数据收集问题,构建了一个复杂的3D动态环境,在多评论家深度确定性梯度算法(MCDDPG)的基础上提出一种基于无人机电量约束、路径长度最小化(MCDDPG-EPM)算法。算法考虑无人机自身电量约束及其在物联网节点间的调度问题,确保无人机在电量供应安全的前提下以较短的路径长度完成数据采集工作。特别地,为了应对动态环境下突发障碍物移动问题,提出信息增强的概念,以降低移动障碍物带来的路径不确定性。仿真结果表明,当物联网节点数为20时,所提算法相较于双延迟深度确定性策略梯度算法(TD3)、传统A*算法和蚁群算法(ACO)分别节省了11.8%、13.2%和15.1%的电量消耗。

关键词: 路径规划, 物联网, 无人机, 多评论家深度确定性梯度算法, 双延迟深度确定性策略梯度算法

Abstract:

Trajectory design is critical for navigation of Unmanned Aerial Vehicle (UAV) as it ensures the provision of safe and reliable travel paths. However, the dynamics of the actual environment significantly amplify the challenges associated with path planning. To address this issue, we design a complex 3D dynamic environment, in which a UAV is employed to collect data from multiple ground nodes of Internet of Things (IoT). Drawing inspiration from the Multi-Critic Deep Deterministic Policy Gradient (MCDDPG), we propose a Multi-Critic DDPG-Energy constrained Path-length-Minimization (MCDDPG-EPM) algorithm, which considers the UAV’s energy constraints and related node scheduling problems, ensuring that the data can be completely collected with short path length and sufficient power. In particular, an information enhancement method is introduced to handle the unpredictability of moving obstacles in dynamic environments. Simulation results demonstrate that compared with the Twin Delayed Deep Deterministic policy gradient algorithm (TD3), traditional A* algorithm, and the traditional Ant Colony Optimization (ACO), the proposed approach achieves energy savings of 11.8%, 13.2%, and 15.1%, respectively, when the number of the IoT nodes is 20.

Key words: trajectory design, internet of things, unmanned aerial vehicle, multi-critic deep deterministic gradient algorithm, twin delayed deep deterministic policy gradient algorithm

中图分类号: