大范围环境中EKF-SLAM问题的改进算法  

Improved EKF-SLAM Algorithm in Large Environments

作　　者：邓伟[1] 梁岚珍[2,1] 浦剑涛[2] 

机构地区：[1]新疆大学电气工程学院,新疆乌鲁木齐830047 [2]北京联合大学自动化学院,北京100101

出　　处：《计算机仿真》2014年第11期345-348,385,共5页Computer Simulation

摘　　要：机器人同步定位与地图构建(SLAM)是指机器人在移动过程中以增量形式创建环境地图并通过所构建地图反复推断自身位置的过程。为实现上述功能,采用传统的扩展卡尔曼滤波(EKF)最优迭代估计方法,在大范围环境条件下,估计误差累积增大,且不能对已构建的环境地图进行更新。提出一种改进算法(KLM-EKF算法),用已知路标的信息对机器人位姿和协方差矩阵进行修正,并创建辅助系数矩阵修正已构建地图,从而实现路标的全局更新。仿真结果表明,在大范围环境中,改进后的算法使机器人自身定位和路标估计误差大幅度降低,并且能够自主地更新已构建地图,有效提高了定位和构图精度。Robot Simultaneous Localization and Mapping （SLAM） is the process by which a mobile robot can build a map of an environment and at the same time this map can be used to compute it＇ s own location. Extended Kalman Filter （EKF） is an optimal iterative estimation method in a wide range of environments, however, its estimated error accumulation increases and it is unable to use the current observed information to update the environmental map which has been built. This paper proposes an improved algorithm （ KLM-EKF）, which can update the pose of robot and covariance matrix by using the information of known landmark, while establish an auxiliary coefficient matrix to achieve the landmarks global update. Simulation results show that, in a large environments, the improved algorithm makes the robot positioning and landmarks estimation error slash, and it can update the built map and effectively improve the accuracy of positioning and composition.

关 键 词：同步定位与地图构建 扩展卡尔曼滤波 大范围环境 辅助系数矩阵 估计误差 

分 类 号：TP242[自动化与计算机技术—检测技术与自动化装置]