噪声相关多传感器系统的微观EKF融合算法

Micro-EKF fusion algorithm for multi-sensor systems with correlated noise

集中式融合滤波对于噪声相关非线性多传感器系统很重要。首先,在扩展Kalman滤波器(EKF)的基础上,利用矩阵求逆引理推导出噪声相关的EKF的一种信息滤波器形式;然后,根据矩阵相似变换理论将其等价分解为具有局部通信的微观滤波器形式。与现有的集中式融合算法相比,新方法保持了相同融合精度的同时,还具备了部分信息滤波器的优良数值计算特点。最后,通过理论分析和计算机仿真相结合的方法来验证了新算法的有效性。

Centralized fusion filtering is important for multi sensor nonlinear systems with correlated noise. An extended Kalman filter （EKF） was combined with matrix inversion in an information form of the EKF with correlated noise （IEKF-CN）. This is then equivalently expressed in consensus form using a micro EKF with local communication based on matrix similarity transformation theory. The computational accuracy of the result is equivalent to existing centralized fusion algorithms while still having the numerical properties of an information filter. The algorithms are validated theoretically and with simulations.