时间频率学报

针对常规卡尔曼滤波应用在GPS/INS组合导航时,由于量测数据出现异常值或系统状态模型不准确而造成的滤波精度下降问题,提出了一种基于新息的自适应卡尔曼滤波算法(AKF)。该算法首先通过卡方检验检测出量测异常值,在量测异常值处调整量测噪声方差阵来抑制滤波发散;在此基础上根据新息协方差的计算值与新息协方差的预测值的粗略比率,调整系统噪声方差阵,从而提高整体滤波精度。通过跑车试验,对本文提出的AKF算法进行了验证。试验结果表明:本文提出的AKF算法较常规卡尔曼滤波算法在经度、纬度误差(均方根)上分别降低了67%,34%,在东向速度、北向速度误差(均方根)上分别降低了47%,38%。从而证明了该算法能有效地抑制由量测异常值导致的状态估计误差,防止滤波发散,提高滤波稳定性。

Aiming at the reduction of the filtering accuracy when the conventional Kalman filter is used in GPS/INS integrated navigation due to outliers of measurement data or inaccurate system state model, an adaptive Kalman filter algorithm based on innovation is proposed. First, the algorithm detects measured outliers by chisquare test, and adjusts the measurement noise variance matrix at the outliers to suppress filtering divergence. Thereby, the rough ratio between the calculated value and the predicted value of the innovation variance matrix is calculated to adjust the system noise variance matrix and then the overall filtering accuracy will be improved. The AKF algorithm proposed in this paper is verified by a vehicle test. The results show that, the proposed AKF algorithm in this paper reduces the longitude errors and latitude errors(root mean square) with 67% and 34% comparing with the conventional Kalman filter algorithm, and reduces the eastward velocity errors and northward velocity errors(root mean square) with 47% and 38%. It is proved that the algorithm can effectively suppress the state estimation error caused by the measured outliers, and further improve the filtering stability.