基于多次测量更新的移动机器人SLAM仿真.PDF

基于多次测量更新的移动机器人SLAM 仿真 1 1 1 2 许亚芳 , 孙作雷 ，曾连荪 ，张波 （1 上海海事大学信息工程学院，上海市 201306 ；2 中国科学院上海高等研究所，上海市 201210 ） 摘 要：同时定位与地图构建(SLAM)是移动机器人实现自治的关键技术，但在用线性化方法处理非 线性系统时，线性化误差的累积会严重影响机器人的估计精度。针对该问题，从费切尔信息角度分 析，提出多次测量更新算法。该算法利用滤波中预测和各次更新阶段状态向量的费切尔信息加权关 系，解算每次测量更新后的状态估计，通过多次将更接近于真实状态的估计值融入观测信息，可得 到较高精度的后验状态估计，降低非线性系统线性化误差的同时提高了机器人定位与构图准确性。 实验部分实现多次测量更新与一次测量更新算法的仿真与对比，结果表明：所提算法的状态估计更 接近于真实值，能有效降低机器人位姿误差和地图估计误差。 关键词：移动机器人；费切尔信息；多次测量更新；线性化误差 中图分类号：TP242 文献标识码：A 文章编号：2015-781X (2002) xx-xxxx-x The Mobile Robot SLAM Simulation with Multi Measurement Update 1 1 1 2 XU Ya -fang , SUN Zuo-lei , ZENG Lian-sun , ZHANG Bo ( 1 Information communication college, Shanghai Maritime University, Shanghai 201306, China; 2 Shanghai Advanced Research Institute, Chinese Academy of Sciences Shanghai 201210, China) Abstract ：Simultaneous Localization and Mapping is one of the key technology in the field of mobile robot. Usually, the accumulation of linearization error will has a severe effect on the robot estimation accuracy when apply linearization method to a non-linear system. To improve the problem, an algorithm named multi measurement update is put forward according to the analysis of Fisher information. In order to compute the state estimation after each measurement update, we make use of the Fisher information weight relationship between prediction variable and update variable. Due to a number of data association with a estimation which is more close to the real data than the former, the algorithm could achieve a more accuracy posterior state . As a result,