一类噪声相关多传感器系统的新型序贯式融合滤波

A novel sequential fusion filtering for multi-sensor system with noise correlations

对于过程噪声与观测噪声一步互相关、各观测噪声一步自相关的多传感器融合滤波问题,提出了一种新的低维序贯式融合滤波算法.基于低阶迭代正交变换的思想提出了解相关的方法,将观测方程经过等价改写去除系统噪声的相关性,然后依据序贯滤波的思想,依次处理到达融合中心的观测信息,进而得到一类实时序贯式融合滤波算法.整个推导过程在线性最小均方误差意义下严格进行,能够实现系统状态的最优融合估计.最后的仿真验证了新算法在处理上述噪声相关问题上的最优性.

A novel low-dimension sequential fusion filtering algorithm is proposed for the multisensor fusion filtering problem with two kinds of system noise correlations：process noise and measurement noise with one-step cross-correlations,the measurement noise with one-step autocorrelations.Based on the low-dimension iterative orthogonal transformation,the method of decorrelation is proposed,by which the measurements can be equivalently transformed as new forms without system noise correlation.Then,combining the idea of sequential filtering,the measurements can be dealt with according to their arriving sequence.Thus,a real time sequential fusion filtering algorithm is obtained.The total deduction is conducted exactly in the sense of linear minimum mean square error（LMMSE）,therefore the fusion estimation of system state is optimal.The final simulation verifies the optimality of the proposed algorithm in dealing with the above noise correlation problem.