摘要
精准定位是实现移动机器人自主导航的先决条件。为解决单一传感器的局限性问题,提出了一种通过扩展卡尔曼滤波(EKF)算法对轮速计和视觉惯性传感器的信息进行融合定位。针对视觉受到遮挡干扰后影响融合定位的问题,提出了基于特征点云个数的改进EKF融合定位算法。该算法通过特征点云个数和方差系数计算函数,动态地更新视觉惯性里程计(VIO)的噪声协方差矩阵,以便消除视觉被遮挡后对系统定位结果的影响。仿真实验结果表明:提出的基于特征点云个数的EKF融合定位和传统EKF融合定位相比,在定位精度上相差无几,但在鲁棒性上大幅提升。实物实验结果进一步验证了所提出的融合定位方法能够有效消除视觉受到遮挡后对定位结果的影响。
Precision localization is a prerequisite for autonomous navigation of mobile robots.To address the limitations of individual sensors,an extended Kalman filtering(EKF)algorithm is proposed to fuse information from wheel encoders and visual-inertial sensor for localization.To solve the problem of vision occlusion affecting the fusion localization,an improved EKF fusion localization algorithm based on the number of feature point clouds is proposed.This algorithm dynamically updates the noise covariance matrix of the visual-inertia odometry(VIO)using a function that calculates the number of feature point clouds and variance coefficients,thereby mitigating the impact of vision occlusion on localization result of the system.Simulation results show that the proposed EKF fusion localization based on the number of feature point clouds has similar localization precision compared to traditional EKF fusion localization but significantly improves robustness.Real experimental results further verify that the proposed fusion localization method can effectively eliminates the influence of visual occlusion on localization results.
作者
胡欢
贾田鹏
张英
HU Huan;JIA Tianpeng;ZHANG Ying(School of Mechnical and Electrical Engineering,Beijing Information Science and Technology University,Beijing 100192,China;School of Modern Post(School of Automation),Beijing University of Posts and Telecommunications,Beijing 100876,China)
出处
《传感器与微系统》
北大核心
2025年第3期125-129,共5页
Transducer and Microsystem Technologies
基金
国家重点研发计划资助项目(2023YFE0122600)。
关键词
多传感器融合定位
扩展卡尔曼滤波
特征点云
噪声协方差矩阵
multi-sensor fusion localization
extended Kalman filtering
feature point cloud
noise covariance matrix
