航空学报 > 2021, Vol. 42 Issue (3): 324278-324278   doi: 10.7527/S1000-6893.2020.24278

基于MCIS技术的相位差变化率单站无源定位

邢怀玺1, 张宇晖1, 陈游1, 周一鹏1, 何文波2   

  1. 1. 空军工程大学 航空工程学院, 西安 710038;
    2. 空军装备部 驻成都第三军事代表室, 成都 610000
  • 收稿日期:2020-05-25 修回日期:2020-07-02 发布日期:2020-09-17
  • 通讯作者: 陈游 E-mail:826476643@qq.com

Single station passive localization with phase difference change rate based on MCIS technology

XING Huaixi1, ZHANG Yuhui1, CHEN You1, ZHOU Yipeng1, HE Wenbo2   

  1. 1. College of Aerospace Engineering, Air Force Engineering University, Xi'an 710038, China;
    2. Third Military Representative Office, Air Force Equipment Department, Chengdu 610000, China
  • Received:2020-05-25 Revised:2020-07-02 Published:2020-09-17

摘要: 针对最大似然估计(ML)方法求解测相位差变化率单站无源定位问题计算量大、定位慢的问题,本文提出一种利用蒙特卡洛重要性抽样技术(MCIS)高精度、低复杂度的估计方法。根据Pincus定理推导出ML问题的近似全局解,利用重要性抽样(IS)技术构建符合高斯分布概率密度(PDF)的重要性函数,作为样本选取的依据,通过逆变换采样获得样本集,统计样本均值直接得到辐射源位置估计结果。MCIS方法简单易实现且运算量低,能够克服传统ML估计多维网格搜索耗时较长的缺陷,而且对目标位置初始估计误差有较低的敏感性。实验结果表明,MCIS算法在相同测量噪声水平下,定位精度优于EKF、NLS算法,有效减小了初始化估计误差对算法定位精度的影响,也进一步讨论分析了算法参数和不同观测条件对定位性能的影响。

关键词: 无源定位, 重要性抽样, 相位差变化率, 最大似然估计, 逆变换采样

Abstract: Aiming at the problems of heavy calculation burden and slow positioning of the Maximum Mikelihood (ML) estimation method for the single-station passive localization via phase difference change rate measurement, this paper proposes a high-precision, low-complexity estimation method using Monte Carlo Importance Sampling (MCIS) technology. According to the Pincus theorem, the approximate global solution of the ML problem is derived. The Importance Sampling (IS) technique is used to construct the importance function that conforms to the Probability Density (PDF) of the Gaussian distribution, which is regarded as the basis for sample selection. The sample set is obtained by inverse transform sampling, and the estimation result of the radiation source position is directly derived by the statistical sample mean. With low sensitivity to the initial estimation error of the target position, the MCIS method is simple and easy to implement with low computational complexity, thereby avoiding the large time consumption of the traditional ML estimation multi-dimensional grid search. Simulation results show that the MCIS algorithm has better positioning accuracy than the Extended Kalman Filter (EKF) and Nonlinear Least Square (NLS) algorithms at the same noise level, and effectively reduces the influence of the initialization estimation error on the positioning accuracy of the iterative algorithm. The influence of the algorithm parameters and different observation conditions on the positioning performance is further discussed and analyzed.

Key words: passive localization, importance sampling, phase difference change rate, maximum likelihood estimation, inverse transform sampling

中图分类号: