融合视觉与运动学的车辆综合定位研究  

Research on integrated vehicle localization by integrating vision and kinematics

作　　者：许男[1] 杨帆[1] 吴晓双 周健锋 XU Nan;YANG Fan;WU Xiaoshuang;ZHOU Jianfeng(National Key Laboratory of Automotive Chassis Integration and Bionics,Jilin University,Changchun 130022,China)

机构地区：[1]吉林大学汽车底盘集成与仿生全国重点实验室,吉林长春130022

出　　处：《长春工业大学学报》2024年第6期500-507,共8页Journal of Changchun University of Technology

基　　金：国家自然科学基金资助项目(52372385)。

摘　　要：针对在城市峡谷、隧道等卫星定位系统失效的情况下,采用单目相机、消费级惯性测量单元和车辆底盘数据作为定位系统输入,为智能车提供定位信息。首先提出动态的轮胎周长模型,并通过对相关参数建立非线性最小二乘问题,求解出待标定的参数优化值,对基于运动学的车辆里程计进行了改进。将改进后里程计的数据融入到定位系统前端的尺度因子初始化过程中,减少车辆启动阶段惯性测量单元噪声产生的不利影响,提高了系统初始化的稳定性。在后端部分增加了车辆的运动学约束,同时进行非线性优化。通过实车实验将其与VINS-Fusion算法结果对比,证明在初始化稳定性与车辆定位精度两方面上的提高。The research purpose of this article is to use monocular cameras,consumer grade Inertial Measurement Units,and vehicle chassis data as inputs to the positioning system in the event of global positioning system failure in urban canyons,tunnels,and other areas,to provide positioning information for intelligent vehicles.In this paper,a dynamic tire circumference model is proposed first,and the optimization values of the parameters to be calibrated are solved by establishing the nonlinear least squares problem for the relevant parameters,and the kinematic based vehicle odometer is improved.Afterwards,the improved odometer data was integrated into the scale factor initialization process of the positioning system front-end,reducing the adverse effects of Inertial Measurement Units noise during vehicle startup and improving the stability of system initialization.In the backend part,for the fusion of multi-sensor data,non-linear optimization is carried out.The vins_wheel proposed in this article was compared with the vins_fusion algorithm through real vehicle experiments,demonstrating the improvement of the algorithm in terms of initialization stability and vehicle positioning accuracy.

关 键 词：同步定位与地图构建 多传感器融合 轮式里程计 车辆运动学 

分 类 号：U463[机械工程—车辆工程]