比例最小偏度单行采样的平方根UKF-SLAM算法

摘要

Due to the distortion of filter gain matrix and high calculation complexity associating with nonlocal effect if symmetric sampling is adopted in the UKF-SLAM algorithm, the UKF-SLAM algorithm adopting square root and single row sampling method based on the proportion and mini-skewness is proposed. The advanced algorithm puts the square root of covariance matrix into iterative computation instead of covariance matrix and adopts the optimized single row sam-pling strategy based on the proportion and mini-skewness. The experimental results show the algorithm can improve the estimation accuracy of feature map and the pose of robot, besides, that can reduce the complexity and increase stability of the algorithm.%对于UKF-SLAM算法所存在的滤波增益矩阵计算失真，采用对称采样计算复杂度相对较高且易产生非局部效应等问题，提出基于比例最小偏度单行采样的平方根UKF-SLAM算法。改进后的算法采用协方差阵的平方根代替协方差阵带入迭代运算，并以比例最小偏度单行采样的方式优化采样策略。仿真结果表明，该算法能够有效地提高机器人位姿以及特征地图的估计精度，并降低了计算复杂度，提高算法的稳定性。