UPF算法在惯导非线性初始对准中的应用

惯性导航系统（INS）的初始对准误差模型通常为非线性的，对于估计惯导误差普遍采用的是扩展卡尔曼滤波算法（EKF），该方法是在一阶泰勒展开的基础上近似得到的，因而误差较大。粒子滤波算法一种新颖的非线性滤波算法，它较传统的EKF算法具有稳定性好，适用范围广的优点。该文首先介绍了作为粒子滤波理论基础的递推贝叶斯估计的基本概念，说明了重要性函数对于粒子滤波器的设计是至关重要的。随后，给出了一种将不敏卡尔曼滤波（UKF）算法作为重要性函数的UPF算法，并提出将其用于静基座条件下的惯导系统非线性初始对准，通过计算机仿真对比了UPF和EKF的估计效果。仿真结果表明，UPF算法较传统的EKF算法对准时间更快，对准精度更高。

Application of Unscented Particle Filter in INS Non-Linear Alignment

The error model of inertial navigation system alignment is usually nonlinear, and the extended Kalman filter is the most widely used estimation algorithm for this error model. Extended Kalman filter is based on first-order Taylor expansion, for which reason, it introduces too much error. Particle filter, which is more stable and has a wider field of application than the traditional EKF, is an novel nonlinear filter algorithm. The principle of recursive Bayesian estimation, as the basis of particle filter, is firstly introduced, and the significance of importance function to the design of particle filter is illustrated. An unscented particle filter （UPF） algorithm whose importance function is unscented Kalman filter is given. Then the UPF is used to estimate the INS alignment on stationary base and the simulation result is compared with EKF＇s. From the simulation result we can see that the UPF is faster in alignment time and more accurate in estimation precision than the EKF in the case of low uncertainties in heading and tilt angle.