基于抗差估计的VIO/UWB自适应组合定位算法  

Adaptive integrated positioning algorithm of VIO/UWB based on robust estimation

作　　者：李雪强 李建胜[1] 王安成 罗豪龙 杨子迪 李凯林 LI Xueqiang;LI Jiansheng;WANG Ancheng;LUO Haolong;YANG Zidi;LI Kailin(Institute of Geospatial Information,Information Engineering University,Zhengzhou 450000,China;Troops 61618,Beijing 100080,China)

机构地区：[1]信息工程大学地理空间信息学院,郑州450000 [2]61618部队,北京100080

出　　处：《中国惯性技术学报》2023年第11期1083-1091,共9页Journal of Chinese Inertial Technology

基　　金：国防科技重点实验室基金(6142403210201);国家自然科学基金(42330113)。

摘　　要：在GNSS信号拒止的室内环境定位中,针对视觉惯性里程计(VIO)在长期运动或无回环等不利环境下产生误差累积偏移以及超宽带(UWB)受非视距影响定位精度难以保证的问题,提出了一种基于抗差估计的UWB辅助视觉惯性自适应组合定位算法。首先,构建UWB与VIO的优化框架,利用UWB定位结果对VIO进行全局约束。其次,在后端优化阶段加入抗差估计,实时调整传感器间的权重值,抑制UWB非视距的影响。最后,在EuRoc数据集和真实场景中进行了实验验证。真实场景实验结果表明,组合算法在非遮挡条件下定位精度相比于VINS_Mono提高75.05%,基于抗差估计的组合算法在遮挡条件下定位精度相比于不加抗差估计的组合算法提高37.53%。In the indoor environment positioning where the global navigation satellite system(GNSS)signal is rejected,the positioning error of visual-inertial odometry(VIO)inevitably accumulates under adverse conditions such as long-term movement or no loop closure,and the positioning accuracy of ultra-wideband(UWB)is difficult to guarantee due to the influence of non-line-of-sight(NLOS).Therefore,an UWB aided visual-inertial adaptive integrated positioning algorithm based on the robust estimation is proposed.Firstly,the optimization framework of UWB and VIO is constructed,and the global constraints on VIO are applied by using the UWB positioning results.Secondly,robust estimation is added in the back-end optimization stage to adjust the weight values between sensors in real time to suppress the influence of NLOS of UWB.Finally,experiments are carried out in EuRoc dataset and real-world scenarios.Experimental results in real-world scenarios show that the positioning accuracy of the combination algorithm is improved by 75.05%compared to VINS_Mono under non-occlusion conditions,and the positioning accuracy of the combination algorithm based on robust estimation is improved by 37.53%compared to the combination algorithm without robust estimation under occlusion conditions.

关 键 词：视觉惯性里程计 超宽带 抗差 自适应 多传感器融合 

分 类 号：V249.328[航空宇航科学与技术—飞行器设计] P228[天文地球—大地测量学与测量工程]