基于轮式里程计辅助视觉的无人车重定位算法
曲萍萍 , 于浩楠 , 于腾丽 , 陈云浩 , 周忠鸿 , 霍文强 , 王尔申
沈阳航空航天大学学报 ›› 2026, Vol. 43 ›› Issue (3) : 61 -68.
基于轮式里程计辅助视觉的无人车重定位算法
Relocalization algorithm for unmanned vehicles based on wheel odometry assisted vision
针对无人车在视觉退化环境中视觉同步定位与地图构建(simultaneous localization and mapping,SLAM)重定位失效的问题,提出一种轮式里程计(wheel odometry,WO)辅助视觉里程计(visual odometry,VO)的自适应协同重定位算法。该算法构建了基于误差状态卡尔曼滤波(error-state Kalman filter,ES-KF)的WO/VO融合模型,提升无人车路径规划的定位精度。针对无人车运行中存在视觉特征点不足导致的难以重定位的问题,提出了一种WO/VO自适应协同重定位算法,通过设计WO/VO系统与WO的切换机制,提升无人车的定位连续性和路径规划的可靠性。利用搭建的无人车平台,在不同情况下对算法进行实验验证与对比。研究结果表明,基于ES-KF的WO/VO融合方法提升了无人车定位性能,绝对轨迹误差(absolute trajectory error,ATE)较单一VO和单一WO分别降低了39.1%和59.2%,路径重合度分别提高了9.2%和11.5%;并在视觉退化场景下实现了WO辅助的自适应协同重定位。ES-KF相较扩展卡尔曼滤波(extended Kalman filter,EKF)与无迹卡尔曼滤波(unscented Kalman filter,UKF)在轨迹精度与位姿估计稳定性方面均表现更优,显著增强了路径融合效果与系统鲁棒性。
To address the issue of SLAM-based visual relocalization failure in unmanned vehicles under visual degradation conditions,an adaptive collaborative relocalization algorithm was proposed,which combined wheel odometry (WO) with visual odometry (VO).This algorithm constructed a WO/VO fusion model based on error-state Kalman filter (ES-KF),enhancing the positioning accuracy of the unmanned vehicle’s path planning.To tackle the challenge of insufficient visual feature points that hinder relocalization during unmaned vehicle operation,an adaptive collaborative relocalization algorithm for WO/VO was proposed.The system included a switching mechanism between WO/VO and WO,improving the continuity of the unmanned vehicle’s localization and the reliability of path planning.Experiments were conducted and compared on the developed unmanned vehicle platform under various conditions.The results demonstrate that the ES-KF-based WO/VO fusion method significantly improves localization performance.The absolute trajectory error (ATE) is reduced by 39.1% and 59.2%,respectively,compared to VO-only and WO-only methods,and path overlap is increased by 9.2% and 11.5%,respectively.The algorithm also successfully achieves WO-assisted adaptive collaborative relocalization in visual degradation scenarios.Compared to the extended Kalman filter (EKF) and the unscented Kalman filter (UKF),the ES-KF demonstrates superior performance in both trajectory accuracy and pose estimation stability,significantly enhancing path fusion effectiveness and system robustness.
| [1] |
曹学伟,袁杰,梁荣光.基于特征点云配准的SLAM重建深度优化方法[J].计算机工程与设计,2025,46(3): 657-664. |
| [2] |
朱代先,魏嘉鑫,刘树林.改进ORB特征匹配的视觉SLAM定位算法[J].激光与光电子学进展,2025,62(12): 168-175. |
| [3] |
纪泽源,于潇颖,付文兴.基于图神经网络特征点匹配的视觉SLAM算法[J].仪器仪表学报,2024,45(9): 34-43. |
| [4] |
陈宗海,裴浩渊,王纪凯, |
| [5] |
马小陆,张睿,郑旗, |
| [6] |
岑志豪,谷德峰,侯文健.可重定位的激光视觉惯性室内定位方法[J].导航定位学报,2024,12(4): 165-173. |
| [7] |
|
| [8] |
谭立超.基于时间同步优化和特征筛选的无人机单目视觉惯导联合定位方法[J].武汉大学学报(工学版),2023,56(2): 237-243. |
| [9] |
包加桐,杨圣奥,朱润辰, |
| [10] |
郭关有,姚仁和,杜白雨, |
| [11] |
张智博,邹喜华,邓果.基于激光与视觉融合的复合机器人高精度导航系统[J].组合机床与自动化加工技术,2023(12): 30-33. |
| [12] |
龚坤,徐鑫,陈小庆, |
| [13] |
邓鹏,罗静.复杂环境下机器人多传感器融合定位方法 [J].电子测量与仪器学报,2023,37 (12): 48-57. |
| [14] |
李玉峰,王君祎,李硕磊.一种基于UWB的高精度室内定位系统设计[J].沈阳航空航天大学学报,2021,38(6): 68-76. |
| [15] |
赵文龙,高建烨,何涛.融合多传感器的自主AGV定位研究[J].现代制造工程,2021(10): 85-90. |
| [16] |
|
| [17] |
|
| [18] |
|
| [19] |
|
| [20] |
蒋林,李云飞,汤勃, |
| [21] |
蔡旭东,王永才,白雪薇, |
| [22] |
|
国家自然科学基金(62173237)
云南省无人自主系统重点实验室开放课题(202501ZD02)
/
| 〈 |
|
〉 |