摘要
对于一个多传感器系统,当其中每个传感器的观测噪声都是独立的有色噪声时,假设系统模型结构已知,而有色噪声的参数和系统噪声方差完全未知时,对其提出了一种两段辨识算法。此算法首先对系统建立相应的自回归滑动平均模型,然后通过递推辅助变量法辨识出自回归参数,接着通过求相关函数构造的若干个线性方程,解出作为未知量的噪声方差,其中部分噪声方差通过算术平均融合得到更精确的估值,融合估值的精确度高于局部估值。该方法可用于GPS导航中的自校正估计。对一个三传感器系统例子用matlab仿真,当步数在2000步以上时估值已稳定收敛于真值,5000步以上稳定在偏差度5%以内,可满足实用要求,从而证明了此辨识方法的有效性。
For a multisensory system, when every single sensor' s measurement noises are independent and colored. Suppose the structure of the model is known, the parameters of the colored noises and all of the noises variance are unknown, present a 2 - stage identification algorithm. First, build the corresponding autoregressive moving average model for the system, the autoregressive parameters can be solved by the recursive instrumental variable algorithm, Then the noises variance can be solved by the linear equations created by the correlation function. The fusion estimation of some noises variance can get by arithmetical average, the precision of fusion estima- tions is higher than local estimations. This algorithm can use for GPS self - tuning navigation. Use matlab to do a simulation for a 3 - sensors example, when the step is 2000 and above, the estimations have stable convergence to the real values, when this step is 5000 and above, the deviation degree of the estimations are stable in 5%. It can meet the practical requirements, proving the validity of the algorithm.
出处
《宜春学院学报》
2015年第6期13-16,共4页
Journal of Yichun University
基金
国家自然科学基金(No.61401101)
东南大学毫米波国家重点实验室开放课题(No.K201401)
阜阳师范学院科技成果孵化基金项目(No.2013KJFH05)
阜阳师范学院自然科学项目(No.2013FSKJ08
No.2013FSKJ14)
阜阳师范学院校自然科学基金重点项目(No.2013FSKJ02ZD)
关键词
多传感器信息融合
有色噪声
两段辨识算法
Multisensory Information Fusion
Colored Noises
Two- stage Identification Algorithm