EKF-SLAM算法的改进及其在Turtlebot的实现

Improved EKF- SLAM Method and Its Implementation on the Turtlebot

针对标准扩展卡尔曼滤波(EKF)在移动机器人同时定位与地图构建(SLAM)过程中计算量大、实时性差、精度低、易受干扰等问题,结合平滑算法和奇异值分解运算,提出了一种基于EKF-SLAM算法的改进ERTSS-SLAM算法。改进ERTSS-SLAM算法使用前向EKF滤波对移动机器人里程计和陀螺仪的位姿信息进行最优估计,再使用标准ERTS平滑器进行后向递推避免发散,同时使用奇异值分解法避免标准EKF滤波产生的求逆运算,有效提高了系统实时性,增强了系统的鲁棒性和定位精度。Turtlebot移动机器人的实验效果证明了该算法在SLAM应用中的高效性和稳定性。

Targeting to tackle those problems like too massive calculation, weak timeliness, low accuracy and robustness, etc. , during the process of the mobile robot' s simultaneous localization and mapping (SLAM) by using the way of the standard extended Kalman filter ( EKF), combining the smoothing algo- rithm and singular value decomposition algorithm, a new SLAM algorithm ( ERTSS - SLAM) is proposed in this paper. By the forward EKF method, the improved ERTSS -SLAM algorithm conducts the optical pose estimation of the mobile robot' s odometer and gyroscopes pose information and then avoid dispersion with the standard ERTS smoother, and at the same time, the inverse calculation is prevented by using the singular value decomposition algorithm, so as to enhance the system' s timeliness, robustness and accura- cy accordingly. Experimental results on the Turtlebot show that the new ERTSS -SLAM algorithm is very efficient and stable during the SLAM application.