噪声相关带偏差线性系统的滤波融合算法

Filtering Fusion Algorithm for Linear Systems with Correlated Noises and Bias

传统线性两阶段Kalman滤波算法无法应对噪声相关情形,导致较低的实际应用性能。针对该问题,以线性系统中状态与测量噪声相关的多传感器偏差估计系统为对象,以基于模型等效变换技术的噪声相关两阶段Kalman滤波器为基本滤波器,分别在序贯分布式和并行式框架下建立两种两阶段Kalman滤波融合算法。其中,序贯分布式融合算法将多个局部两阶段Kalman滤波器的估计结果以序贯加权的形式进行融合;并行式融合算法分别对偏差滤波估计和无偏差滤波估计进行融合,再利用线性方程将融合后的结果进行组合,得到状态估计值。仿真结果表明:相比于两阶段Kalman滤波器和序贯分布式两阶段Kalman滤波融合估计器,并行式两阶段Kalman滤波融合估计器在滤波估计精度上具有更高的性能。

The traditional linear two-stage Kalman filtering algorithm can not cope with the situation with correlated noises,and its practical application performance is low.For this problem,the multi-sensor bias estimation system in which state noise is correlated to measurement noise is taken as the object,a two-stage Kalman filter with correlated noises based on equivalent transformation technique of model is used as basic filter,two kinds of two-stage Kalman filtering fusion algorithms are established in sequential distributed and parallel framework,respectively.The sequential distributed fusion algorithm fuses the estimates of multiple local two-stage Kalman filter in a sequential weighted form,while the parallel fusion algorithm fuses the estimates of the bias filter and the bias-free filter separately;and then the linear equation is used to combine the fused results to obtain the state estimation.The simulation results show that the parallel two-stage Kalman filter fusion estimator has higher performance in filtering estimation accuracy than the two-stage Kalman filter and the sequential distributed two-stage Kalman filter fusion estimator.