摘要
针对惯性/重力/重力梯度组合导航EKF算法中,随机线性化方法不考虑重力场固有属性以及验前随机变量的不确定因素,从而导致真实验后均值和协方差携带较大误差,滤波性能损失甚至发散问题,依据重力、重力梯度本身物理特性,提出基于地球重力场球谐模型的随机线性化方法。进而,为保证滤波实时性,建立了适用于并行计算的地球重力场模型线性化矩阵运算公式,并通过GPU并行方案提高了模型线性化过程的计算效率。某海域仿真实验结果表明:利用重力场球谐模型线性化方法进行EKF滤波匹配,相较于九点拟合法和双二次曲面拟合法的导航定位精度均提高了37.5%以上,可以有效削弱线性化误差,避免滤波发散,提高系统导航定位精度,且线性化耗时优于0.02 s,满足匹配导航实时性需求。
In the EKF algorithm of inertial/gravity/gravity gradient integrated navigation,the random linearization method does not consider the inherent attributes of the gravity field and the uncertainties of the prior random variables,which leads to large errors in the mean and covariance after the real experiment,loss of filtering performance and even divergence.According to the physical characteristics of gravity and gravity gradient,a random linearization method based on the spherical harmonic model of the earth gravity field is proposed.Furthermore,in order to ensure the real-time filtering,the linearization matrix calculation formula of the earth gravity field model suitable for parallel calculation is established,and the calculation efficiency of the linearization process of the model is improved by the GPU parallel scheme.The simulation results of a certain sea area show that the EKF filter matching by using the linearization method of spherical harmonic model of gravity field improves the navigation and positioning accuracy by more than 37.5%compared with the nine point fitting method and the biquadratic surface fitting method.It can effectively weaken the linearization error,avoid filter divergence,improve the navigation and positioning accuracy of the system,and the linearization time is better than 0.02 s,which meets the real-time requirements of matching navigation.
作者
黄炎
李姗姗
谭勖立
王傲明
范雕
付林威
HUANG Yan;LI Shanshan;TAN Xuli;WANG Aoming;FAN Diao;FU Linwei(PLA Strategic Support Force Information Engineering University,Zhengzhou 450001,China)
出处
《中国惯性技术学报》
EI
CSCD
北大核心
2022年第3期328-335,共8页
Journal of Chinese Inertial Technology
基金
国家自然科学基金(42174007,42174013)。
