摘要
粒子滤波在组合导航系统非线性非高斯条件下的滤波估计中获得广泛关注,但捷联惯导误差模型维数较高,直接应用粒子滤波会带来维数灾难。设计了用于SINS/GPS组合导航的卡尔曼/粒子组合滤波算法,采用卡尔曼滤波和粒子滤波分别对系统的线性和非线性状态进行估计,降低粒子滤波器状态维数,避免维数灾难。采用系统残差采样法的规则化粒子滤波器,有效缓解粒子贫化问题,并减少计算负担。仿真结果表明卡尔曼/粒子组合滤波方法的估计性能与粒子滤波相当,但计算复杂度前者要低得多。
Particle filter is a best way for the state estimation of highly nonlinear integrated navigation systems with non-Gaussian uncertainties. Since the error model of SINS has high dimensions, traditional particle filter would bring hard computation. A new Kalman/Particle mixed filter used on SINS/GPS integrated navigation system was proposed. The new method divides the system into two sub-models, one is linear, the other one is nonlinear, and then implement Kalman filter and particle filter separately. Residual-systematic resampling and regularized algorithms were involved to decrease particle filter’ particles collapse weakness and hard computation. The simulation results show that their performance is almost equal, but the computation complexity of the Kalman/particle filter is much lower than traditional particle filter.
出处
《系统仿真学报》
CAS
CSCD
北大核心
2009年第1期220-223,共4页
Journal of System Simulation
基金
863计划(2006AA12Z305)
关键词
捷联惯导
组合导航
粒子滤波器
卡尔曼/粒子组合滤波器
strap-down inertial navigation system
integrated navigation system
particle filter
kalman/particle filter
