基于头脑风暴算法的FastSLAM 2.0算法  

FastSLAM 2.0 algorithm based on brain storm optimization

作　　者：朱代先[1] 王明博 刘树林[3] 郭苹 Zhu Daixian;Wang Mingbo;Liu Shulin;Guo Ping(College of Communication&Information Engineering,Xi’an University of Science&Technology,Xi’an 710054,China;College of Energy Engineering,Xi’an University of Science&Technology,Xi’an 710054,China;College of Electrical&Control Engineering,Xi’an University of Science&Technology,Xi’an 710054,China)

机构地区：[1]西安科技大学通信与信息工程学院,西安710054 [2]西安科技大学能源学院,西安710054 [3]西安科技大学电气与控制工程学院,西安710054

出　　处：《计算机应用研究》2021年第12期3629-3633,共5页Application Research of Computers

基　　金：国家自然科学基金资助项目(51774235)。

摘　　要：针对FastSLAM 2.0算法粒子权值退化与粒子多样性丧失导致机器人定位建图精度下降的问题,提出了基于头脑风暴算法改进FastSLAM 2.0算法。通过头脑风暴算法替换FastSLAM 2.0算法重采样过程,首先将重要性采样后的粒子权值作为头脑风暴算法中个体评判的适度值,根据适度值大小差异完成K-means聚类操作;其次对聚类后的集合进行变异操作,并取消头脑风暴算法中个体选择操作,从而实现改进头脑风暴算法替代FastSLAM 2.0算法重采样过程,缓解粒子的贫化现象,增加粒子多样性,最终实现对机器人定位建图精度的提升。在机器人定位建图实验中,对比经典FastSLAM 2.0算法和基于遗传算法改进FastSLAM 2.0算法,提出的算法定位精度最高,相较于经典FastSLAM 2.0算法,提出算法定位精度提升了63%,稳定性提升了55%。Aiming at the problem that the degradation of particle weight and the loss of particle diversity in FastSLAM 2.0 algorithm led to the decline of robot positioning and mapping accuracy,this paper proposed an improved FastSLAM 2.0 algorithm based on brain storm optimization.Firstly,it took the particle weight after importance sampling as the moderation value of individual evaluation in the brainstorming algorithm,and completed K-means clustering operation according to the difference of the moderation value.Secondly,it mutated the set after clustering to cancel the individual selection operation in the brainstorm algorithm.In this way,the improved brainstorm algorithm could replace the resampling process of particle filter,effectively alleviate the particle dilution phenomenon,increase the particle diversity,and finally improved the positioning and mapping accuracy of the rescue robot.In the robot positioning mapping experiment,compared with the classic FastSLAM 2.0 algorithm and the improved FastSLAM 2.0 algorithm based on genetic algorithm,the proposed algorithm has the highest positioning accuracy.Compared with the classic FastSLAM 2.0 algorithm,the positioning accuracy of the proposed algorithm is improved by 63%,and the stability is improved by 55%.

关 键 词：机器人 同时定位与建图 FastSLAM 2.0 头脑风暴算法 粒子权值退化 粒子贫化 重采样 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]