基于高斯分布重采样的Rao-Blackwellized粒子滤波-控制与决策.PDFVIP

第31 卷 第12 期 控 制 与 决 策 2016 年 12 月 Vol. 31 No. 12 Control and Decision Dec. 2016 文章编号: 1001-0920 (2016) 12-2299-06 DOI: 10.13195/j.kzyjc.2015.1227 基于高斯分布重采样的Rao-Blackwellized 粒子滤波SLAM 算法 张 毅, 郑潇峰, 罗 元, 庞冬雪 (重庆邮电大学信息无障碍工程与机器人技术研发中心，重庆400065) 摘 要: 针对移动机器人同时定位与地图构建中RBPF-SLAM 算法因粒子匮乏而导致栅格地图估计不精确问题, 提 出一种基于高斯分布重采样的RBPF-SLAM 算法. 首先, 根据粒子权重大小对重采样粒子进行排序; 然后, 在重采样 中利用高斯分布分散高权重粒子得到新粒子, 从而保证粒子多样性, 避免粒子匮乏, 保证栅格地图的精确构建. 实验 结果表明了所提出算法的有效性, 同时也证明该算法能在粒子数减少的条件下保持可靠的估计, 有效地减少了计算 量. 关键词: 移动机器人；同时定位与地图构建；Rao-Blackwellized 粒子滤波；高斯分布重采样 中图分类号: TP242.6 文献标志码: A SLAM algorithm with Gaussian distributed resampling Rao- Blackwellized particle filter ZHANG Yi, ZHENG Xiao-feng, LUO Yuan, PANG Dong-xue (Research and Development Center of Information Accessibility Engineering and Robotics，Chongqing University of Posts and Telecommunications，Chongqing 400065，China. Correspondent ：ZHANG Yi ，E-mail ：zhangyi@) Abstract: For the estimation problem that the RBPF-SLAM algorithm used in mobile robots suffers from sample impoverishment in grid mapping, a Gaussian distributed resampling(GDR) based RBPF-SLAM algorithm is proposed. Firstly, the improved algorithm sorts particles according to the weight size. Furthermore, Gaussian distributed resampling is applied to disperse the high-weight particles so as to generate new particles. By using GDR, particle diversity can be maintained and sample impoverishment can be avoided. Thus accurate grid mapping is guaranteed. Experimental results show the effectiveness of the proposed algorit