改进重采样的移动机器人SLAM算法  

SLAM algorithm for mobile robot based on improved resampling

作　　者：张廷军 郭毅锋 黄丽敏[2] ZHANG Ting-jun;GUO Yi-feng;HUANG Li-min(School of Electrical and Information Engineering,Guangii University of Science and Technology,Liuzhou 545006,China;HTC VIVEDU School of Technlogy,Guangii Uiiversity of Scieicz and Technology,Liuzhou 545006,China)

机构地区：[1]广西科技大学电气与信息工程学院,广西柳州545006 [2]广西科技大学宏达威爱科技学院,广西柳州545006

出　　处：《计算机工程与设计》2019年第11期3276-3281,共6页Computer Engineering and Design

基　　金：国家自然科学基金项目(51407038);广西创新驱动发展专项科技重大专项基金项目(桂科AA17204062)

摘　　要：在基于Rao-Blackwellized粒子滤波的移动机器人同时定位与地图构建(RBPF-SLAM)算法中,针对重采样过程导致粒子多样性降低问题,提出部分粒子免疫优化重采样方法。根据粒子权重将粒子划分为稳定粒子和不稳定粒子,对稳定粒子直接存入抗体记忆序列,对不稳定粒子通过计算抗原与抗体的亲和力与排斥力进行克隆变异操作,从中优选新粒子补充到抗体记忆序列,提高粒子多样性。实验结果表明,该算法能够有效提高机器人状态估计精度,保证算法实时性。In the simultaneous localization and mapping algorithm of mobile robot based on Rao-Blackwellized particle filter(RBPF-SLAM),aiming at the problem of particle diversity reduction caused by resampling process,apartial particle immune optimal resampling method was proposed.The particles were divided into stable particles and unstable particles according to their weights.The stable particles were stored directly into the antibody memory sequence.Unstable particles were cloned and mutated by computing the affinity and repulsive forces of antigens and antibodies.New particles were selected from them and added to the antibody memory sequence.The particle diversity was improved.Experimental results show that the improved algorithm can effectively improve the accuracy of robot state estimation and guarantee the real-time performance of the algorithm.

关 键 词：粒子滤波 同时定位与地图构建 重采样 克隆变异 免疫优化 

分 类 号：TP242.6[自动化与计算机技术—检测技术与自动化装置]