摘要
针对现有导航解算中将模型误差作为过程噪声,并假设为高斯白噪声来处理,可能会引起较大的估计误差,甚至导致滤波器发散的问题。文章以提高导航解算精度为目的,在吸收非线性预测滤波和粒子滤波优点的基础上,提出一种新的非线性模型预测Unscented粒子滤波算法。该算法在建立系统模型时顾及了模型误差,对模型误差进行实时估计,并利用该估计值对包含模型误差的非线性、非高斯系统模型进行修正,从而提高了导航解算精度。将提出的算法应用于捷联惯性导航(SINS)/合成孔径雷达(SAR)组合导航系统进行仿真验证,并与预测滤波和Unscented粒子滤波进行比较,结果表明,提出的新算法不但滤波性能明显优于预测滤波和Unscented粒子滤波,而且能提高导航解算精度。
In navigation calculations, model error is assumed as some kind of noise for Gaussian white noise to process, thus causing large state estimation error of nonlinear filtering system and even divergence. We present a new NMPUPF algorithm based on predictive filtering and unscented particle filtering to improve the accuracy of nav- igation calculations. Section 1 of the full paper explains our NMPUPF algorithm, which we believe is new and effec- tive. Section 1 gives a fairly detailed 6-step procedure for using our new algorithm. The core of the 6-step procedure consists of： （1） initialization, （2） calculating state estimation which includes the model error, （3） calculating es- timation of systematical output vector, （4） calculating model error, （ 5 ） correcting nonlinear system model, （6） recycle calculations. Simulation results, presented in Figs. 1 through 4, and their analysis demonstrate preliminarily that the proposed algorithm is indeed effective for filtering calculations based on nonlinear and non-Gaussian system model, thus improving significantly the filter performance and integrated navigation calculation accuracy.
出处
《西北工业大学学报》
EI
CAS
CSCD
北大核心
2012年第5期734-738,共5页
Journal of Northwestern Polytechnical University
基金
国家自然科学基金(61174193)
陕西省自然科学基金(NBYU0004)资助
关键词
模型误差
预测滤波
Unscented粒子滤
模型预测Unscented粒子滤波
algorithms, calculations, computer simulation, errors, inertial navigation systems, Kalman filters, mathematical models, nonlinear systems, real time systems, state estimation, synthetic aperture radar
model error, Nonlinear Model Predictive Unscented Particle Filtering（NMPUPF）
作者简介
作者简介：高怡（1978-），女，西北工业大学博士研究生，主要从事控制理论与控制工程研究。
