中国惯性技术学报
中國慣性技術學報
중국관성기술학보
JOURNAL OF CHINESE INERTIAL TECHNOLOGY
2012年
6期
678-686
,共9页
魏伟%秦永元%张晓冬%张亚崇
魏偉%秦永元%張曉鼕%張亞崇
위위%진영원%장효동%장아숭
卡尔曼滤波%Sage-Husa算法%自适应%数据扰动环境
卡爾曼濾波%Sage-Husa算法%自適應%數據擾動環境
잡이만려파%Sage-Husa산법%자괄응%수거우동배경
Kalman filter%Sage-Husa algorithm%adaptive%data disturbance environment
在衍生于卡尔曼滤波的INS/GPS组合导航自适应算法中,Sage-Husa算法是针对不确定的系统噪声协方差阵Q阵和量测噪声协方差阵R阵的经典算法.文中推导了自适应Q阵和R阵,在自适应Q阵中补充了Sage-Husa算法遗失的一些量.为了尽可能多保留信息,构造新方法使得Q阵非负定和R阵正定,取消了易导致估计偏差的均值估计,修改了Sage-Husa算法中使初始噪声方差的作用很快趋于零的设计思想,可控制固定噪声方差的比例.在3种数据扰动环境下,利用动态仿真计算分析和比较了相关算法发现,同时对Q阵和R阵做自适应估计的算法,其抵抗数据扰动的效果要优于固定Q阵只对R阵做自适应估计、或固定R阵只对Q阵做自适应估计的算法,这个结论不同于一些文献的观点.新改进的 Sage-Husa算法对自适应Q阵和R阵可分别调节,有较好的适应性和稳定性.
在衍生于卡爾曼濾波的INS/GPS組閤導航自適應算法中,Sage-Husa算法是針對不確定的繫統譟聲協方差陣Q陣和量測譟聲協方差陣R陣的經典算法.文中推導瞭自適應Q陣和R陣,在自適應Q陣中補充瞭Sage-Husa算法遺失的一些量.為瞭儘可能多保留信息,構造新方法使得Q陣非負定和R陣正定,取消瞭易導緻估計偏差的均值估計,脩改瞭Sage-Husa算法中使初始譟聲方差的作用很快趨于零的設計思想,可控製固定譟聲方差的比例.在3種數據擾動環境下,利用動態倣真計算分析和比較瞭相關算法髮現,同時對Q陣和R陣做自適應估計的算法,其牴抗數據擾動的效果要優于固定Q陣隻對R陣做自適應估計、或固定R陣隻對Q陣做自適應估計的算法,這箇結論不同于一些文獻的觀點.新改進的 Sage-Husa算法對自適應Q陣和R陣可分彆調節,有較好的適應性和穩定性.
재연생우잡이만려파적INS/GPS조합도항자괄응산법중,Sage-Husa산법시침대불학정적계통조성협방차진Q진화량측조성협방차진R진적경전산법.문중추도료자괄응Q진화R진,재자괄응Q진중보충료Sage-Husa산법유실적일사량.위료진가능다보류신식,구조신방법사득Q진비부정화R진정정,취소료역도치고계편차적균치고계,수개료Sage-Husa산법중사초시조성방차적작용흔쾌추우령적설계사상,가공제고정조성방차적비례.재3충수거우동배경하,이용동태방진계산분석화비교료상관산법발현,동시대Q진화R진주자괄응고계적산법,기저항수거우동적효과요우우고정Q진지대R진주자괄응고계、혹고정R진지대Q진주자괄응고계적산법,저개결론불동우일사문헌적관점.신개진적 Sage-Husa산법대자괄응Q진화R진가분별조절,유교호적괄응성화은정성.
Among the INS/GPS integrated navigation adaptive algorithms derived from Kalman filter, the Sage-Husa algorithm is the classic method for the uncertain system noise covariance matrix Q and measurement noise covariance matrix R . In this paper, the adaptive matrix Q and matrix R were deduced, some values missed by the Sage-Husa algorithm were made up in the adaptive matrix Q . In order to save the information as much as possible, a new method was constructed to make matrix Q non-negative definite and matrix R positive definite. The mean estimations which could easily induce the estimation departure were cancelled. The design idea in the Sage-Husa algorithm, i.e. the effect of the initial noise variance is quickly approaching zero, was modified, so the proportion of the fixed noise variance could be controlled. Under the environments with 3 kinds of data disturbance, the dynamic simulation computation was used to compare and analyze related algorithms. It is found that, the effect of resisting data disturbances of the algorithm which simultaneously adaptively estimates matrix Q and matrix R is better than the algorithm adaptively estimating matrix R with the fixed matrix Q , or the algorithm adaptively estimating matrix Q with the fixed matrix R , and this conclusion is different from some papers’ viewpoint. The modified Sage-Husa algorithm, which can respectively adjust the adaptive matrix Q and matrix R , has better flexibility and stability.