地理科学进展 ›› 2021, Vol. 40 ›› Issue (9): 1516-1527.doi: 10.18306/dlkxjz.2021.09.007

• 无人机运行监管 • 上一篇    下一篇

基于人群密度风险的无人机动态路径规划研究

焦庆宇(), 陈新锋*(), 郑志刚, 柏艺琴, 刘艳思, 张正娟, 孙龙妮   

  1. 中国民航科学技术研究院,北京 100028
  • 收稿日期:2020-12-29 修回日期:2021-03-30 出版日期:2021-09-28 发布日期:2021-11-28
  • 通讯作者: * 陈新锋(1964— ),女,北京人,教授级高工,主要从事适航维修管理研究。E-mail: chenxf@mail.castc.org.cn
    * 陈新锋(1964— ),女,北京人,教授级高工,主要从事适航维修管理研究。E-mail: chenxf@mail.castc.org.cn
  • 作者简介:焦庆宇(1995— ),男,河南新乡人,硕士生,主要从事空中交通运输规划与管理研究。E-mial: jiaoqingyu1125@sina.com
  • 基金资助:
    国家重点研发计划项目(2017YFB0503005);中国科学院重点部署项目(ZDRW-KT-2020-2-1);天津科技计划项目智能制造专项(Tianjin-IMP-2018-2)

Dynamic path planning of unmanned aerial vehicle based on crowd density prediction

JIAO Qingyu(), CHEN Xinfeng*(), ZHENG Zhigang, BAI Yiqin, LIU Yansi, ZHANG Zhengjuan, SUN Longni   

  1. China Academy of Civil Aviation Science and Technology, Beijing 100028, China
  • Received:2020-12-29 Revised:2021-03-30 Online:2021-09-28 Published:2021-11-28
  • Supported by:
    National Key Research and Development Program of China(2017YFB0503005);The Key Research Program of Chinese Academy of Sciences(ZDRW-KT-2020-2-1);Tianjin Intelligent Manufacturing Project(Tianjin-IMP-2018-2)

摘要:

为解决无人机数量的快速增长导致其对地面尤其是城市内运行风险的提升,提高无人机的运行效率,减少无人机对地面人群造成的威胁,需基于人群密度对无人机制定特定的路径规划。然而,现阶段对无人机进行路径规划时仍以静态人口统计数据作为地面风险的分析依据,未能根据人群密度的时空变化特征对无人机进行实时路径规划。论文首先分析城市路网人群密度时空数据特征;其次,利用卷积神经网络对不同区域的人群密度进行预测;最后,根据已预测的人群密度数据,利用改进A*算法对无人机进行实时路径规划及风险评估。使用该模型对北京上空无人机路径进行规划,结果显示,无人机运行风险降低了76%,可为无人机交通管理系统实时路径规划功能的建立提供理论参考。

关键词: 无人机空中交通管理, 风险运行, 卷积神经网络, 人群密度, 路径规划

Abstract:

With the rapid growth of the number of unmanned aerial vehicles (UAVs), in order to assess the risk of people on the ground especially in cities, improve the operational efficiency of UAV, and reduce the threat of UAV to the crowd, it is necessary to make a specific path planning for UAVs based on the crowd density. However, static demographic data are still used as the ground risk analysis basis for the path planning of UAV, and real-time path planning of UAV according to the spatial-temporal characteristics of crowd density are often not performed. This study first analyzed the characteristics of urban road network crowd density based on the spatiotemporal data. Second, the convolutional neural network-long short-term memory (CNN-LSTM) combined model (C-Snet model) was established to predict the population density in different urban areas. Finally, the improved A* algorithm was used for real-time path planning and risk assessment of UAV according to the predicted crowd density data. The results show that the risk of UAV operation is reduced by 76%, which can provide a theoretical reference for the development of real-time path planning function of UAV traffic management system.

Key words: UAV traffic management, risk assessment, CNN, crowd density, path planning