航空学报 > 2019, Vol. 40 Issue (8): 322969-322969   doi: 10.7527/S1000-3893.2019.22969

基于核主成分分析的空域复杂度无监督评估

张瞩熹1,2, 朱熙1,3, 朱少川1,4, 张明远1,5, 杜文博1,5   

  1. 1. 北京航空航天大学 综合交通大数据应用技术国家工程实验室, 北京 100083;
    2. 中国人民解放军32751单位;
    3. 北京航空航天大学 前沿科学技术创新研究院, 北京 100083;
    4. 北京航空航天大学 交通科学与工程学院, 北京 100083;
    5. 北京航空航天大学 电子信息工程学院, 北京 100083
  • 收稿日期:2019-02-26 修回日期:2019-03-06 出版日期:2019-08-15 发布日期:2019-08-26
  • 通讯作者: 朱熙 E-mail:zhuxi@buaa.edu.cn
  • 基金资助:
    国家杰出青年科学基金(61425014);国家优秀青年科学基金(61722102);国家自然科学基金(61671031)

Unsupervised evaluation of airspace complexity based on kernel principal component analysis

ZHANG Zhuxi1,2, ZHU Xi1,3, ZHU Shaochuan1,4, ZHANG Mingyuan1,5, DU Wenbo1,5   

  1. 1. National Engineering Laboratory of Big Data Application Technologies of Comprehensive Transportation, Beihang University, Beijing 100083, China;
    2. Unit 32751 Force of PLA, China;
    3. Research Institute of Frontier Science, Beihang University, Beijing 100083, China;
    4. School of Transportation Science and Engineering, Beihang University, Beijing 100083, China;
    5. School of Electronic and Information Engineering, Beihang University, Beijing 100083, China
  • Received:2019-02-26 Revised:2019-03-06 Online:2019-08-15 Published:2019-08-26
  • Supported by:
    National Science Fund for Distinguished Young Scholars (61425014); National Science Foundation for Excellent Youth (61722102); National Natural Science Foundation of China (61671031)

摘要: 空域复杂度评估作为衡量空域运行态势、管制员工作压力的关键手段,是运行调控的基础。由于影响因素众多,不同因素间耦合关联复杂,且标定样本很难获取,空域复杂度的准确评估被公认为航空领域的一个挑战性问题。提出了一种空域复杂度的无监督评估方法。首先通过核主成分分析挖掘原始样本各维度的非线性耦合关系,准确提取能够最大化复杂度评估信息量的主成分,进一步设计了可按需定制的主成分聚类方法,实现了无监督条件下空域复杂度的准确评估,为空域划分、流量管理提供了有效的技术支撑。

关键词: 空域复杂度, 空域运行态势, 无监督, 核主成分分析, 降维, 聚类

Abstract: Airspace complexity evaluation is a key means to measure the airspace operational situation and the controller workload, providing the basis for the operation optimization. Its accurate evaluation is a challenging problem in the aviation domain due to numerous influencing factors, the complex correlations between factors, and the high difficulty of collecting labelled samples. This paper proposes an unsupervised evaluation method for airspace complexity. Firstly, the kernel principal component analysis is utilized to mine the nonlinear correlations in different sample dimensions, and extract several principal components in which the airspace complexity information is maximized. Furthermore, the principal component clustering which can be customized according to user requirements is designed. The proposed method achieves accurate complexity evaluation capacity under the unsupervised condition, providing effective technical support for air traffic management like airspace configuration and traffic management.

Key words: airspace complexity, airspace operation situation, unsupervised learning, kernel principal component analysis, dimension reduction, clustering

中图分类号: