基于空间域划分的分布式SLAM算法

Distributed simultaneous localization and mapping algorithm based on partition of space-region

针对同步定位与地图构建(simultaneous localization and mapping,SLAM)中状态量高维时变的问题,本文通过综合集中式和分布式实现结构的各自优势,提出了一种基于空间域划分的分布式SLAM算法。该算法依据两个路标点与机器人连线之间的夹角,将整个空间域中的路标点进行区域划分,保证每个子空间域内含有两个不共线的路标点,并将每个空间域内的路标点组合构建观测模型,采用分布式无味粒子滤波器进行机器人位姿的估计,而采用联邦Kalman滤波完成对路标点的估计,并通过设计各子滤波器中粒子分布的调整方式改善了系统在动态重构过程的精度和稳定性。最后,通过实际数据的仿真试验证明所提算法具有更好的实时性和滤波精度。

To solve the problem of the simultaneous localization and mapping （SLAM） under the complex circumstance, a distributed algorithm of the SLAM based on partition of space-region is proposed considering the respective advantages of centralized configuration and distributed structure. The region is formed according to the angle between two landmarks and the robot, which is designed in case of the collinearity between two landmarks. The landmarks in each region are combined to establish the corresponding observation model. Be sides, the position of the robot is obtained by applying the distributed unscented particle filter and the positions of the landmarks are estimated simultaneously by employing the Kalman filter. Meanwhile, the accuracy and the stability are improved through constructing the adjustment of particle distribution during the dynamic reconfiguration process. Eventually, the better real-time capability and filter accuracy of the proposed SLAM algorithm are proved through simulation experiments which are supported by actual data.