基于SVD的自适应无迹H_(∞)滤波定位算法研究  

Adaptive Unscented H_(∞)Filtering Localization Algorithm Based on SVD

作　　者：伊鹏 柏建军 时丙新 陈炳旭 沈秀强 YI Peng;BAI Jianjun;SHI Bingxin;CHEN Bingxu;SHEN Xiuqiang(College of Automation,Hangzhou Dianzi University,Hangzhou Zhejiang 310018,China;PetroChina Jidong Oilfield Oil and Gas Gathering and Transportation Company,Tangshan Hebei 063200,China;Zhejiang Zhengtai Zhongzi Control Engineering Co.,Ltd.,Hangzhou Zhejiang 310018,China)

机构地区：[1]杭州电子科技大学自动化学院,浙江杭州310018 [2]中国石油冀东油田油气集输公司,河北唐山063200 [3]浙江正泰中自控制工程有限公司,浙江杭州310018

出　　处：《传感技术学报》2024年第10期1726-1733,共8页Chinese Journal of Sensors and Actuators

基　　金：国家自然科学基金项目(61773146,61973102,U22A2044);广东省智能决策与协同控制重点实验室资助项目。

摘　　要：针对移动机器人噪声模型不确定性导致定位算法鲁棒性弱、精度低的问题,提出一种基于奇异值分解(Singular Value Decomposition,SVD)的自适应无迹H_(∞)滤波定位算法。该算法利用无迹H_(∞)滤波融合多传感器数据估计移动机器人位姿,并通过自适应调节滤波器参数γ,提高了移动机器人的定位精度。同时为了提高算法的鲁棒性,采用SVD分解代替常规Cholesky分解,避免了误差协方差矩阵在数值迭代过程中出现负定的情况。实验结果表明:相较于扩展H_(∞)滤波和粒子滤波算法,基于SVD分解的自适应无迹H_(∞)滤波定位算法具有精度高、鲁棒性强的优势。Targeting at the problems of low robustness and accuracy of the localization algorithm of mobile robot caused by the indefi niteness of the noise model,an adaptive unscented H_(∞)filtering localization algorithm based on singular value decomposition(SVD)is proposed.Firstly,the pose of the mobile robot is estimated by combining the data of the odometry and the LiDAR.Then,the filter pa rameters are adaptively adjusted to improve the accuracy of the localization.Meanwhile,singular value decomposition(SVD)is used to improve the robustness of the numerical calculation and the negative definiteness of the error covariance is avoided,which is always en countered by the conventional Cholesky decomposition.Finally,some experimental results are given to show the advantages of the algo rithm proposed in this paper.

关 键 词：移动机器人定位 多传感器融合 自适应 无迹H_(∞)滤波 SVD 

分 类 号：TP242[自动化与计算机技术—检测技术与自动化装置]