针对扩展卡尔曼滤波在递推过程中状态协方差可能失去正定性,从而引起滤波发散现象的问题,在滤波过程中用协方差平方根代替协方差进行迭代计算,保证其正定性。对一组异类传感器(红外传感器和雷达)观测数据采用加权最小二乘法进行融合,然后用平方根扩展卡尔曼滤波对融合的数据进行滤波。蒙特卡罗仿真结果表明,平方根扩展卡尔曼滤波和加权最小二乘数据融合方法可以保证滤波精度,并且能有效抑制滤波发散。
Aiming at the problem of the filter divergence induced by the possible losing of positive definition of the covarianee matrix of state in extended Kalman filter iteration, the eovariance square-root matrix (CSRM), instead of the covariance matrix(CM), is used in the filter iteration. The CSRM can effectively keep its positive definition. The observed data from the heterogeneous sensors (infrared sensor and Radar) is fused by using the weighted least square, and then the result of the measurement fusion is filtered by the square-root extended Kalman filter. The Monte-Carlo simulation shows that the square-root extended Kalman filter and the weighted least square can guarantee the filtering precision and reduce the filter divergence effectively.