基于卡尔曼滤波和粒子滤波器级联模型的静基座惯导初始对准算法及仿真

传统静基座初始对准主要采用扩展卡尔曼滤波技术。扩展卡尔曼滤波器本质上要求系统近似线性。当近似线性要求得不到满足，会产生很大的偏差，大大降低对准精度，甚至会发散。粒子滤波是一种新出现的滤波技术，对模型不作线性限制，非常适于解决非线性问题，估计精度大大高于传统扩展卡尔曼滤波器。但是，要求维数不能太高，否则会产生计算灾难问题。惯导误差模型的维数较高，这使得粒子滤波技术无法实际应用于初始对准中。本文通过对静基座误差方程进行分析，提出了一种级联模型。将原有模型分解为级联的两个子模型，每个子模型的状态变量维数都很低，然后对两个子模型分别应用卡尔曼滤波器和粒子滤波器进行滤波处理。实验仿真结果表明，这种基于卡尔曼滤波器和粒子滤波器级联模型的算法降低了计算量，大大提高了初始对准的精度，具有重要的现实意义。

Algorithm and simulation of stationary base inertial alignment based on cascaded model of EKF and PF

Traditional method for stationary base initial alignment is generally Extended Kalman Filtering(EKF),which,in essence,requires that the system model is approximately linear.When the requirement cannot be met, great error will be generated,even leading to divergence.Particle Filtering(PF) is a new tech for filtering,which is suitable for the nonlinear problems with much higher estimating precision than EKF,but only in the case when the model dimension is not too high.Since the model of inertial navigation error has high dimensions,traditional particle filter cannot be applied to its initial alignment.The authors analyzed the stationary-base inertial error model and put forward a novel algorithm: first,dividing the system into two cascaded sub-models with lower dimensions,and then implementing extended Kalman filtering and particle filtering alternately.Simulation result showed that the new algorithm improved the alignment precision and decreased the computation cost greatly.It is of great significance in practice.