[1]陈 欣,赵之硕,吴佳琦,等.基于贝叶斯-UKF分层融合的智能轮椅导航系统研究[J].机械与电子,2026,44(08):51-57.
　CHEN Xin,ZHAO Zhishuo,WU Jiaqi,et al.Research on an Intelligent Wheelchair Navigation System Based on Bayesian-UKF Hierarchical Fusion[J].Machinery & Electronics,2026,44(08):51-57.
点击复制

基于贝叶斯-UKF分层融合的智能轮椅导航系统研究()
分享到：

《机械与电子》[ISSN:1001-2257/CN:52-1052/TH]

卷:
44
期数:
2026年08期
页码:
51-57
栏目:
机器人技术
出版日期:
2026-08-25

文章信息/Info

Title:
Research on an Intelligent Wheelchair Navigation System Based on Bayesian-UKF Hierarchical Fusion
文章编号:
1001-2257(2026)08-0051-07
作者:
陈 欣; 赵之硕; 吴佳琦; 郑卓韬
华东理工大学机械与动力工程学院,上海 200237
Author(s):
CHEN Xin; ZHAO Zhishuo; WU Jiaqi; ZHENG Zhuotao
(School of Mechanical and Power Engineering,East China University of Science and Technology,Shanghai 200237,China)
关键词:
智能轮椅; 多传感器融合; 贝叶斯推理; 无迹卡尔曼滤波; 同时定位与地图构建
Keywords:
intelligent wheelchair; multi-sensor fusion; Bayesian inference; unscented Kalman filter; simultaneous localization and mapping
分类号:
TP242
文献标志码:
A
摘要:
针对智能轮椅室内导航中因传感器异构、系统非线性及动态性导致的精度不足问题,提出基于贝叶斯推理与无迹卡尔曼滤波(UKF)的3层多传感器融合导航方法。感知层以对数似然比递推框架融合激光雷达、超声波与红外传感器的异步观测,输出栅格占据概率图;状态估计层以UKF-SLAM 替代EKF-SLAM,通过无迹变换传播非线性统计量并自适应调整观测协方差;决策层设计带双反馈的并行代价地图,与上层构成完整闭环。仿真结果表明:所提方法绝对轨迹误差(ATE)较EKF-SLAM 降低约39%,交并比(IoU)提升约10百分点,动态场景成功率达91%;实物建图误差较Gmapping降低约60%,CPU 占用率满足嵌入式实时要求。
Abstract:
To address the accuracy degradation in indoor navigation of intelligent wheelchairs caused by sensor heterogeneity,system nonlinearity,and environmental dynamics,a three-layer multi-sensor fusion navigation method based on Bayesian inference and the Unscented Kalman Filter (UKF) is proposed.In the perception layer,a recursive log-likelihood ratio framework fuses asynchronous observations from LiDAR,ultrasonic,and infrared sensors to generate a grid occupancy probability map.In the state estimation layer,UKF-SLAM replaces EKF-SLAM,propagating nonlinear statistics through the unscented transform while adaptively adjusting the observation covariance.In the decision layer,a parallel cost map with dual feedback is designed,forming a complete closed loop with the upper layers.Simulation results show that,compared with EKF-SLAM,the proposed method reduces the Absolute Trajectory Error (ATE) by approximately 39%,improves the Intersection over Union (IoU) by about 10 percentage points,and achieves a success rate of 91% in dynamic scenarios.In physical experiments,the mapping error is reduced by roughly 60% compared with Gmapping,and the CPU occupancy meets embedded real time requirements.

参考文献/References:

[1] Kovacevic N,Ahmad I,Clark A,et al.An overview of smart wheelchair technologies[C]∥Health Informatics and Medical Systems and Biomedical Engineering (CSCE 2024).Cham:Springer,2025:73-83. [2] 潘美君,陶镛汀,朱红梅.基于多传感器融合的自主移动机器人导航方法[J].传感器与微系统,2025,44(10):156-159. [3] 李飞,喻自烘,李昊.多传感器融合的移动机器人室内定位方法研究[J].机械设计与制造,2025(11):106-108. [4] 陈路,谢维斯,谭杰,等.面向移动机器人的多传感器紧耦合导航定位方法[J].电子科技大学学报,2026,55(1):109-115. [5] Zhang Hongkai,Du Liang,Bao Sheng,et al.LVIO Fusion: tightly-coupled LiDAR-visual-inertial odometry and mapping in degenerate environments[J].IEEE Robotics and Automation Letters,2024,9(4):3783-3790. [6] Lee J,Komatsu R,Shinozaki M,et al.Switch-SLAM: switching-based LiDAR-inertial-visual SLAM for degenerate environments[J].IEEE Robotics and Automation Letters,2024,9(8):7270-7277. [7] 孙海波,童紫原,唐守锋,等.基于卡尔曼滤波与粒子滤波的SLAM研究综述[J].软件导刊,2018,17(12):1-3. [8] Julier S J,Uhlmann J K.Unscented filtering and nonlinear estimation[J].Proceedings of the IEEE,2004,92(3):401-422. [9] 刘艳,程诚,裴少婧.鲁棒自适应无迹卡尔曼滤波的 SLAM 算法[J].电光与控制,2019,26(8):12-16. [10] 赵玏洋,闫利.移动机器人SLAM 位姿估计的改进四元数无迹卡尔曼滤波[J].测绘学报,2022,51(2):212-223. [11] 方正,佟国峰,徐心和.基于贝叶斯滤波理论的自主机器人自定位方法研究[J].控制与决策,2006,21(8):841-847. [12] Zhu Jun,Li Hongyi,Zhang Tao.Camera,LiDAR,and IMU based multi-sensor fusion SLAM:a survey[J]. Tsinghua Science and Technology,2024,29(2):415-429. [13] Manjooran D C,Selvarajan N P.GUI-controlled autonomous wheelchair navigation with robust sensor fusion and recovery[C]∥2025 International Conference on Robotics and Mechatronics(ICRM).Kollam, India:IEEE,2025:1-6. [14] 阮晓钢,陈晓,朱晓庆.基于多重信息增益的移动机器人探索策略[J].北京工业大学学报,2023,49(9):990-998. [15] 邵磊,张飞,刘宏利,等.融合改进A* 算法与动态窗口法的移动机器人路径规划[J].天津理工大学学报,2024,40(1):71-76. [16] 王军晓,黄美琴,冯建涵,等.基于改进A* 算法和动态窗口法的移动机器人路径规划[J].高技术通讯,2026,36(2):123-136.

备注/Memo

备注/Memo:
收稿日期:2026-06-08 基金项目:国家级大学生创新创业训练计划项目(202510251095) 作者简介:陈 欣 (1989-),女,福建松溪人,博士,实验师,研究方向为机械设计及制造;赵之硕 (2004-),男,河南郑州人,研究方向为机器人控制、嵌入式开发。
更新日期/Last Update: 2026-08-28