移动机器人SLAM位姿估计的改进四元数无迹卡尔曼滤波EI北大核心CSCD

在全自主运动控制的移动机器人系统中,自身位姿的估计和校正对于移动机器人的运动至关重要。卡尔曼滤波是解决移动机器人同步定位与地图构建(SLAM)常用方法。相较于卡尔曼滤波,无迹卡尔曼滤波(UKF)无须对复杂的非线性函数进行雅可比矩阵运算。本文基于无迹卡尔曼滤波,根据先验协方差的平方根选择sigma点,计算协方差以及加权均值。用四元数表示姿态,将四元数矢量转换为旋转空间进行矩阵运算,在此基础上设计了一种位姿估计算法——基于四元数平方根的无迹卡尔曼滤波(QSR-UKF)算法。试验将EKF、QSR-UKF、SR-UKFEKF 3种算法的位姿估计结果进行仿真分析,并通过相关定量指标进行了描述,验证了本文算法的有效性。

Advanced quaternion unscented Kalman filter based on SLAM of mobile robot pose estimation

In automatic motion controlled mobile robot system,the estimation and correction of its own pose is very crucial for the motion of robot.Kalman filter is a classical method to solve the problem of simultaneous localization and mapping(SLAM)in robot system.Compared with Kalman filter,unscented Kalman filter(UKF)uses nonlinear model directly,avoids operation of Jacobian matrix of complex nonlinear function.In this paper,based on the unscented Kalman filter,sigma points are selected by square root decomposition of prior covariance,and then weighted mean and covariance are calculated.In addition,Quaternion is used to represent attitude of robot,and quaternion vector is converted to rotation space for matrix operation.According to the characteristic of square root decomposition and quaternion vector,a quaternion square root unscented Kalman filter is proposed.By comparing the robot poses estimation results on quaternion square root unscented Kalman filter(QSR-UKF),square root unscented Kalman filter(SR-UKF)and extended Kalman filter(EKF),the simulation results show that the proposed QSR-UKF method is effective.