基于逆协方差交叉融合的多传感器非线性滤波器

Multi-sensor nonlinear filter based on inverse covariance intersection fusion

为了解决互协方差未知的多传感器非线性系统的融合估计问题,针对非线性系统提出了基于序贯逆协方差交叉和并行逆协方差交叉的两种容积卡尔曼融合估计算法。各个子系统通过容积卡尔曼滤波器得到滤波估计,分别利用序贯逆协方差交叉融合算法和并行逆协方差交叉融合算法对各子系统局部估计进行融合。两种算法可有效地避免求解高维的权系数凸优化问题,降低了计算负担。当传感器数目很多时,并行逆协方差交叉融合算法因其多层并行结构可以显著节约融合时间。最后仿真结果证明了算法的有效性。

Two cubature Kalman fusion estimation algorithms based on sequential inverse covariance intersection and parallel inverse covariance intersection for nonlinear systems are presented,in order to solve the fusion estimation problem of multi-sensor nonlinear systems with unknown cross-covariance.The cubature Kalman filter is used to obtain the filter estimates of each subsystem,and then the sequential inverse covariance intersection fusion algorithm and the parallel inverse covariance intersection fusion algorithm are used to fuse the local estimates of each subsystem.The two algorithms can effectively avoid the high-dimensional convex optimization problem of solving weighted coefficients,and greatly reduce the computational burden.When the number of sensors is particularly large,the parallel inverse covariance intersection fusion algorithm can significantly reduce the fusion time because of its multi-layer parallel structure.Finally,the simulation results prove the effectiveness of the algorithms.