摘要
通过对传统的基于扩展卡尔满滤波器(EKF)的SLAM算法的介绍,总结出传统方法的缺陷,即算法复杂,用时长,无法实现在线计算.为解决传统SLAM算法的缺陷,介绍了一种基于Rao-Blackwellized粒子滤波器的FastSLAM方法.该方法将SLAM问题分解为对机器人姿态和路标在地图中的位置的递归算法,其时间消耗与路标的数量成对数关系,计算量小,用时短.经过以Hebut-II机器人为平台的实验,结果表明,FastSLAM算法是可行的.
Summarizes the disadvantage of the traditional SLAM based on EKF through the introduction. Such as complexity and lasting long time, unable to achieve calculation online. To solve this problem, this paper presents FastSLAM, which is an algorithm that recursively estimates the full posterior distribution over robot pose and landmark locations, and the time it requires scales logarithmically with the number oflandmarks in the map. Both emulate and real-world (Hebut II robot) experiments were carried out. The results show that, FastSLAM algorithm is a doable in both emulate and real-world environment.
出处
《河北工业大学学报》
CAS
北大核心
2009年第3期37-40,共4页
Journal of Hebei University of Technology
基金
国家863计划资助(2006AA04Z221)
河北省自然科学基金资助(E2006000030)
河北省教育厅科学研究计划(2005222)