摘要
针对陀螺存在随时间累积漂移误差的特性造成卡尔曼滤波数据发散的问题,提出以陀螺与磁强计组合为测量手段,小波变换和卡尔曼滤波相结合的方法。该方法由卡尔曼滤波器获得较为准确的测量信息,对陀螺和磁强计的信息分别在尺度3下进行分解,比较每层的细节系数和近似部分的系数获得误差信号,结合BP神经网络建立误差模型。当测量数据发散时,选用模型的估计值修正卡尔曼滤波测量值。实验结果比传统卡尔曼滤波算法精度提高了2°,能有效提高姿态测量的精度。
To prevent data divergence of the gyro caused by drift error in Kalman filtering,a method was put forward which taking the gyro and magnetometer as measurement instruments by combining the Kalman filtering with wavelet transform.The relatively accurate measurement information was obtained through the Kalman filter,and the information of the gyro and the magnetometer was decomposed under the three dimensions.The error signal was obtained by comparing each level＇s coefficient in detail part and the approximate part,and the error model was built up based on BP neural network.When the data was diverging,the estimated value of the model was used to replace the measured value of the Kalman filter.The simulation result indicated that the measurement accuracy of the method is 2° higher than the traditional Kalman algorithm,and the attitude measurement precision can be improved effectively.
出处
《电光与控制》
北大核心
2011年第8期47-50,共4页
Electronics Optics & Control
基金
国防重点实验室基金(9140C1204041009)
关键词
姿态测试
卡尔曼滤波
小波变换
BP神经网络
attitude measurement
Kalman filtering
wavelet transform
BP neural network
