摘要
在卫星拒止环境中,通常需要集成多种类型的传感器以实现无人车的高精度自主导航。然而,目前的多传感器融合方法普遍是基于定制型的解决方案,不具备对异类异步导航信息的即插即用能力。因此,提出了一种基于惯性/轮速里程计(IMU/ODO)预积分的多传感器"即插即用"因子图融合方法。考虑到轮速里程计标度因数、惯性传感器与轮速里程计之间安装误差的影响,推导并构建了完整的惯性导航系统和轮速里程计的预积分模型,利用预积分信息将无人车运动状态与多源异步量测信息进行关联并采用固定频率进行解算,从而实现了一种即插即用式的因子图导航融合与参数优化方法。实际跑车测试结果表明,所提出方法可有效兼容多类导航传感器信息。同时,相比基于卡尔曼滤波的多源信息融合方法,惯性/轮速里程计/视觉里程计定位精度提升60%以上。
In GNSS-denied environments,it is usually necessary to integrate multiple types of sensors to achieve high-precision autonomous navigation of unmanned ground vehicle.However,current multisensory fusion methods are generally based on customized solutions,and do not have the plug-and-play capability for heterogeneous and asynchronous navigation information.Therefore,a multisensory plug-and-play factor graph fusion method based on IMU/odometer preintegration is proposed.The complete IMU and odometer preingration model is derived in considering the effect of the scale factor of odometer,installation errors between IMU and odometer.The preintegration measurements are utilized to associate the motion states of the unmanned ground vehicle with the asynchronous measurements and the calculation is conducted at a fixed frequency,thereby realizing a plug-and-play factor graph fusion and parameters optimization method.The field experimental test results illustrate the proposed method can be effectively compatible with multiple types of navigation sensors information.Meanwhile,the positioning accuracy of IMU/ODO/VO is improved by more than 60%compared with the multisensory fusion method based on Kalman filter.
作者
白师宇
赖际舟
吕品
岑益挺
王炳清
黄凯
BAI Shiyu;LAI Jizhou;LV Pin;CEN Yiting;WANG Bingqing;HUANG Kai(Nanjing University of Aeronautics and Astronautics,Nanjing 210096,China;Shanxi Baocheng AviationInstrument Corporation,Baoji 721000,China)
出处
《中国惯性技术学报》
EI
CSCD
北大核心
2020年第5期624-628,637,共6页
Journal of Chinese Inertial Technology
基金
国家自然科学基金(61973160,61703207)。
