一种基于RBUKF滤波器的SLAM算法

Algorithm of SLAM Based on RBUKF

同时定位与建图(SLAM)是智能机器人实现真正自治的必要前提,是一个比单独研究定位或者建图更加困难的课题。该文将基于SUT变换的RBUKF滤波器应用于平面静态环境下的同时定位与建图算法,它能够在同样计算复杂度的情况下,避免基于扩展卡尔曼滤波器(EKF)SLAM算法由于线性化误差大导致滤波器发散,从而出现建图错误的缺点。基于公共数据集的实验表明该方法估计的最终地图比EKF的方法精度高。

Simultaneous Localization And Mapping（SLAM） is a necessary prerequisite to make robot autonomous, which is a harder research topic than localizing or mapping. A Rao-Blackwellised Unscented Kalman Filter（RBUKF） based SLAM method is presented which uses the Scaled Unscented Transformation（SUT） to sample the Sigma points for robot operating in plain static environment. With the same computing complexity, RBUKF can avoid linearization error introduced in the Extended Kalman Filter（EKF） filter, which can induce the final map error. The experimental result of the method based on the public dataset is better than the EKF based method according to the precise of the final estimated map.