摘要
纯卡尔曼滤波在系统或观测噪声不符合假设前提下,滤波将出现发散,而纯H_∞滤波虽然稳健,但是精度不高,所以,比较常用的方式是两种滤波形式相结合的H_2/H_∞混合滤波。本文提出了一种不需对噪声作过多建模,对两种滤波的增益矩阵进行加权求和的混合滤波形式,基于该形式,推导了加权系数的求法,并基于该算法,应用在GPS/DR组合导航中,进一步,为了简化计算,提高运算效率,提出了更符合工程应用的稳态H_2/H_∞滤波方法,仿真结果表明,H_2/H_∞混合滤波即具有卡尔曼滤波的同等精度,又具有H_∞滤波的滤波稳健性,综合时间耗费、滤波精度、滤波稳健性三个指标,稳态H_2/H_∞滤波更符合工程应用。
Divergence of Pure Kalman filtering will happen while observation or system noise don' t accorded with hypothesis,and pure H∞ filtering, hasn't enough precision although robustness, so, the popular style is the mixed H2/H∞ filtering, this paper aims at the problem, put forward a kind of filtering way which needn' t create model, bring out a kind of mixed filtering mode using a weighted com- bination of the HE and H~ gains, deduce the weighted coefficient based on this idea, apply the algorithm to the GPS/DR integrated navi- gation system, more, in order to predigest computation and improve efficiency, deduce the steady pure H∞ filtering, the simulation indicate that H2/H∞ filtering not only has the same precision as kalman filtering ,bus also has the same robustness as pure H∞ filtering, the resuits prove that steady H2/H∞ filtering is more practical in time- consumption, precision and robustness.
出处
《信号处理》
CSCD
北大核心
2009年第2期280-284,共5页
Journal of Signal Processing
基金
国防预研项目(0308XG0900)资助