卫星辅助的无人机视觉惯性自主定位方法  

A Vision-inertial Autonomous Localization Method for UAVs with Satellite Assistance

作　　者：高治明 罗子娟 李雪松[2,3] 李春雨 王佳楠 GAO Zhiming;LUO Zijuan;LI Xuesong;LI Chunyu;WANG Jianan(School of Aerospace Engineering,Beijing Institute of Technology,Beijing 100081;National Key Laboratory of Information Systems Engineering,Nanjing 210007;The 28th Research Institute of China Electronics Technology Group Corporation,Nanjing 210007;School of Mechanical Engineering,Tsinghua University,Beijing 100084)

机构地区：[1]北京理工大学宇航学院,北京100081 [2]信息系统工程重点实验室,南京210007 [3]中国电子科技集团公司第二十八研究所,南京210007 [4]清华大学机械工程学院,北京100084

出　　处：《导航与控制》2024年第2期41-50,共10页Navigation and Control

摘　　要：无人机在复杂环境中的自主定位是无人机智能化和应用的重要前提,但也面临着多种挑战,如传感器噪声、动态初始化等。为了解决这些问题,提出了一种基于卫星辅助的无人机视觉惯性自主定位方法,实现了无人机位置、速度、姿态和偏差等状态的实时估计。采用基于IMU预积分的闭式解模型,实现了视觉惯性导航系统(VINS)的动态初始化,有效降低了计算量和初始化时间;对GPS-IMU时间偏移进行建模,并对外部时间偏移进行在线校准,解决了全局位姿测量信息的时延问题,实现了全局位姿信息的异步更新;在单目惯性导航传感器和四旋翼无人机实验平台上进行了户外飞行实验,验证了所提方法的有效性和鲁棒性。实验结果表明,该方法在有间歇性GPS信号复杂环境中的定位精度达到了2.2 m,显著优于传统VINS方法的4.8 m,体现了该方法的高效性和鲁棒性,为无人机在复杂环境中的自主定位提供了一种创新的解决方案,为无人机的智能化和应用开辟了新的领域。Autonomous localization of UAVs in complex environments is an important prerequisite for the intelligence and application of UAVs,but it also faces various challenges,such as sensor noise,dynamic initialization.To solve these problems,a satellite assisted vision-inertial autonomous localization method for UAVs is proposed,which realizes the realtime estimation of the UAVs’position,velocity,attitude and bias and other states.A closed-form solution model based on IMU pre-integration is adopted,which achieves dynamic initialization of the visual-inertial navigation system(VINS),effectively reducing the computation and initialization time.The GPS-IMU time offset is modeled,and online calibration of the external time offset is performed,which solve the time delay of the global pose measurement information,and realize the asynchronous update of the global pose information.Outdoor flight experiments are conducted on a monocular inertial navigation sensor and a quadcopter UAV experimental platform,which verify the effectiveness and robustness of the proposed method.The experiment results show that the positioning accuracy of the proposed method in complex environments with intermittent GPS signals reaches 2.2 m,significantly better than the traditional VINS method of 4.8 m,reflecting the efficiency and robustness of the proposed method.It provides an innovative solution for the autonomous localization of UAVs in complex environments,and opens up new fields for the intelligence and application of UAVs.

关 键 词：无人机状态估计 视觉惯性导航系统 未知环境下自主定位 多源异构传感器 

分 类 号：TP242.6[自动化与计算机技术—检测技术与自动化装置]