Journal of South China University of Technology(Natural Science Edition)

• Low-Altitude Traffic System • Previous Articles     Next Articles

Research on Real-time Quantitative Assessment of UAV Aerial Risk Based on Nonparametric Error Modeling

LI Chenglong 1,2 YUE Yiyang2 ZHANG Xuejun1  ZHENG Yuan3  LUO Yucong4 WEI Peng5   

  1. 1. School of Electronic and Information Engineering, Beihang University, Beijing 100191, China;

    2. School of Air Traffic Management, Civil Aviation Flight University of China, Guanghan 618307, China;

    3. College of Air Traffic Management, Civil Aviation Flight University of China, Chengdu 641419 China;

    4. Xinjin Branch Courts, Civil Aviation Flight University of China, Chengdu 611431, China;

    5. Department of Mechanical and Aerospace Engineering, Geroge Washington University, Washington DC 20052, USA

  • Published:2025-11-07

Abstract:

Aiming at the high collision risk problem faced by large unmanned aerial vehicles and manned aircraft in the coordinated operation of complex airspace, this paper proposes a collision probability estimation method that integrates historical trajectory data and non-parametric error modeling to break through the limitations of traditional models in terms of error characterization ability and dynamic adaptability. At the method level, in order to solve the problem of discontinuity of trajectory data and difficulty in explicitly identifying the error direction, an error extraction mechanism based on B-spline fitting and heading angle decomposition is constructed, and the optimal fitting parameters are determined through ablation experiments. Kernel density estimation is used to flexibly model the navigation error, thereby getting rid of the limitation of fixed distribution assumption on modeling accuracy, and an analytical model that can realize probability estimation at any time is proposed, which is verified by Monte Carlo simulation. Experimental results show that the proposed method shows higher accuracy and stability in collision probability estimation than traditional models. Further sensitivity analysis is carried out in combination with random forest method, and relative distance is identified as the dominant factor affecting risk, which provides theoretical support for risk monitoring and strategy optimization for highly dynamic airspace.

Key words: UAV, mid-air collision, kernel density estimation, Monte Carlo simulation, sensitivity analysis