基于自适应稳健最小p范数估计的轨道确定方法研究

发布时间：2018-10-30 07:00

【摘要】：当测轨数据误差不服从正态分布时,传统的最小二乘(LSE)轨道确定方法将不是最优的.为了获得高精度的定轨结果,一种可行的策略是采用基于最小p范数(Lp)的轨道确定方法.通过分析Lp估计的相关性质,得出普通Lp估计不具有良好的抗差性的结论.为抑制模型误差和异常值的影响,提出了基于数据深度加权的稳健最小p范数估计方法,并证明了相关性质,得出了其崩溃点可以达到1/2的结论.最后,通过残差分析和矩估计法自适应估计相关参数,使得估计达到最大效率.以天基空间目标监视系统为背景进行了仿真试验.结果表明,当观测数据存在系统误差或异常值时,或者当目标动力学模型存在误差或者天基观测平台存在系统误差时,即使观测数据服从正态分布,LSE也不是最优的,在这种意义下自适应稳健Lp估计轨道确定方法比传统轨道确定方法更加稳健,定轨精度也更高.
[Abstract]:When the error of orbit data is not satisfied with the normal distribution, the traditional least square (LSE) orbit determination method will not be optimal. In order to obtain high precision orbit determination results, a feasible strategy is to adopt the orbit determination method based on minimum p-norm (Lp). By analyzing the related properties of Lp estimators, it is concluded that ordinary Lp estimators do not have good robustness. In order to suppress the influence of model errors and outliers, a robust minimum p-norm estimation method based on data depth weighting is proposed, and the correlation properties are proved, and the conclusion that the collapse point can reach 1 / 2 is obtained. Finally, the residual error analysis and the moment estimation method are used to estimate the parameters adaptively to maximize the efficiency of the estimation. The simulation experiment is carried out on the background of space-based space target surveillance system. The results show that when there are systematic errors or outliers in the observed data, or when there are errors in the dynamic model of the target or in the space-based observation platform, LSE is not optimal even if the observed data is normally distributed. In this sense, the adaptive robust Lp estimation orbit determination method is more robust than the traditional orbit determination method, and the orbit determination accuracy is higher.
【作者单位】：国防科学技术大学九院;北京航天飞行控制中心载人航天与深空探测飞行力学实验室;
【基金】：国家自然科学基金(60974124) 载人航天与深空探测飞行力学实验室开放基金资助项目
【分类号】：P135

【参考文献】