基于多模型方法的自适应卡尔曼滤波

针对一类观测噪声统计特性未知的离散时间系统设计一种多模型自适应卡尔曼滤波器。基于多个不同的固定观测噪声协方差阵建立多个固定模型卡尔曼滤波器,将多个固定模型卡尔曼滤波器和一个常规自适应卡尔曼滤波器共同组成多模型自适应卡尔曼滤波器。针对每一个滤波器建立一个基于输出误差的指标切换函数,每一个采样时刻将指标切换函数取得最小值的滤波器的状态估计值切换为系统的当前状态估计值。仿真结果表明,与常规的自适应滤波器相比,此方法可以极大地改善滤波器的滤波效果。

Adaptive Kalman Filter Based on Multiple Model Method

A kind of multiple model adaptive Kalman filter (MMAKF) was set up for the discrete time system without the statistic knowledge of measured noise. Multiple fixed Kalman filters according to the system with different fixed noise covariance matrices and a conventional adaptive Kalman filter were used to form a multiple model Kalman filter. An index switching function in the form of the function of output error of different Kalman filter was designed. At every sample time, the state value of the Kalman filter which gives the minimum value of index switching function will be switched as the estimated value of the state of system. From the simulation, it can be seen that the method proposed can improve the result of conventional Kalman filter greatly.