基于高斯分布的多层无迹卡尔曼滤波算法

• 作 者: 刘江 王玉金 段建雷 叶松庆
• 机 构: 中国科学院重庆绿色智能技术研究院自动推理与认知研究中心,重庆邮电大学计算机科学与技术学院,
• 出 处: 控制与决策
• 发表日期: 2016年04期
• 关键词: 无迹 卡尔曼滤波 无迹卡尔曼滤波 多层采样 卡尔曼滤波算法 高斯分布 高阶矩 高斯滤波 精度 采样算法

在传统无迹卡尔曼滤波(UKF)中对其估计精度和计算效率起关键作用的是采样算法,即构造具有权重的样本点.研究表明,带权样本点匹配随机变量的阶矩越高滤波的精度越高,如多项式无迹卡尔曼滤波(PUKF),但通常此类算法的复杂度过高甚至难以求解.为此,基于高斯分布结合高阶矩匹配与无迹卡尔曼滤波线性扩张方法(LUKF),提出一种兼顾效率和精度的高斯滤波离线算法.实验结果表明,所提出算法拥有比UKF更高的估计精度和比PUKF更好的计算效率.

In the traditional unscented Kalman filter (UKF), its estimation accuracy and computational efficiency play a key role in the sampling algorithm, which is to construct weighted sample points.Research shows that the higher the order of the random variables with weighted sample points filtering , Such as polynomial unscented Kalman filter (PUKF), but usually the complexity of this algorithm is too high or even difficult to solve.Therefore, based on the Gaussian distribution combined with the high-order moment matching and unscented Kalman filtering linear expansion method (LUKF), a new off-line Gaussian filter algorithm with both efficiency and accuracy is proposed.The experimental results show that the proposed algorithm has higher estimation accuracy than UKF and better computational efficiency than PUKF.