超宽带多锚点弹性辅助的视觉/惯性紧耦合导航  

Ultra-wideband multi-anchor resilient-assisted vision/inertial tightly coupled navigation

作　　者：罗豪龙 李建胜[1] 徐志涛 王力[1] 张世田[4] 李广云[1] LUO Haolong;LI Jiansheng;XU Zhitao;WANG Li;ZHANG Shitian;LI Guangyun(School of Geospatial Information,Information Engineering University,Zhengzhou 450001,China;State Key Laboratory of Geo-Information Engineering,Xi’an 710054,China;School of Computer Science and Engineering,Central South University,Changsha 410083,China;The 22nd Research Institute,China Electronics Technology Group Corporation,Qingdao 266107,China)

机构地区：[1]信息工程大学地理空间信息学院,河南郑州450001 [2]地理信息工程国家重点实验室,陕西西安710054 [3]中南大学计算机学院,湖南长沙410083 [4]中国电子科技集团公司第二十二研究所,山东青岛266107

出　　处：《光学精密工程》2025年第3期452-465,共14页Optics and Precision Engineering

基　　金：国家自然科学基金资助项目(No.42071454);地理信息工程国家重点实验室基金资助项目(No.SKLGIE2023-M-2-2);电波环境特性及模化技术重点实验室基金资助项目(No.6142403180204)。

摘　　要：视觉惯性UWB松耦合导航方法需要提前构建UWB坐标系统和获取UWB锚点位置,导致它在实际应用中耗费大量时间和精力。针对单锚点辅助的视觉惯性紧耦合导航方法,由于单个UWB锚点只提供了一维距离约束信息,对系统的精度和稳定性的提升有限。为了降低系统的配置难度和提高系统的定位性能,提出了一种无需提前获取UWB锚点位置、具有普遍适用性的UWB多锚点弹性辅助视觉/惯性导航紧耦合方法。首先,基于即插即用的思想提出了一种UWB多锚点在线异步位置估计方法,以确保UWB锚点信息的及时利用。其次,采用弹性因子图优化框架建立视觉惯性UWB紧耦合的因子图模型,从而实现视觉/惯性/UWB多锚点的紧耦合导航。最后,利用开源VIRAL数据集和真实世界采集的数据集对所提出的方法进行了验证。实验结果表明:相较于先进的VINS-MONO和VIR-SLAM开源算法,本文方法的定位精度分别平均提升了30.61%和23.19%。这表明所提出的UWB多锚点弹性辅助视觉/惯性紧耦合导航方法具有更高的定位精度和更强的稳定性。Due to the limitations of distance measurement from a single ultra-wideband(UWB)anchor,which provides only one-dimensional information,significant enhancement in system accuracy and robust⁃ness is constrained.To address this issue,a UWB multi-anchor resilient-assisted vision/inertial tightly coupled navigation method was proposed,which obviates the necessity of pre-obtaining UWB anchor positions and demonstrates universal applicability.Initially,based on the plug-and-play concept,an online asynchronous po⁃sitioning estimation method was introduced to effectively and timely utilize UWB anchor information.Subse⁃quently,a resilient factor graph optimization framework was employed to construct a tightly coupled factor graph model for vision-inertial UWB fusion,facilitating tightly coupled navigation that integrates information from vision,inertial sensors,and multiple UWB anchors.The proposed method was validated using both the open-source VIRAL dataset and datasets collected from real-world environments.Experimental results indi⁃cate that,in comparison to state-of-the-art open-source algorithms VINS-MONO and VIR-SLAM,the posi⁃tioning accuracy of the proposed method improved by an average of 30.61%and 23.19%,respectively.These findings underscore that the UWB multi-anchor resilient-assisted vision/inertial tightly coupled naviga⁃tion method provides enhanced positioning accuracy and improved robustness.

关 键 词：UWB多锚点 异常值检测 锚点位置估计 因子图优化 紧耦合导航 

分 类 号：TP242[自动化与计算机技术—检测技术与自动化装置] P258[自动化与计算机技术—控制科学与工程]