基于ROS的惯性导航和视觉信息融合的移动机器人定位研究

发布时间：2018-03-10 13:38

  本文选题：移动机器人定位　切入点：惯性导航　出处：《哈尔滨工业大学》2017年硕士论文　论文类型：学位论文

【摘要】：移动机器人的定位是自主导航研究的基础,也是移动机器人领域的重要研究课题。本论文的主要研究内容是基于ROS操作系统,将惯性导航和视觉传感器信息融合,解决移动机器人自主定位的问题。目前,单目SLAM存在初始化的尺度问题和追踪的尺度漂移问题;惯性导航数据稳定,但积累误差更为严重。这两种技术的融合可以将单目视觉SLAM的高精度与惯性导航数据的稳定性结合起来,取长补短,达到满足定位精度的目的。具体研究内容如下:第一,研究了捷联惯性导航系统的基本定位理论,包括惯性导航坐标系,基本旋转理论和基于四元数算法的姿态解算的知识以及航位推算的理论。为进一步研究和应用奠定了基础。并进行实验测试了惯性导航的定位精度;第二,研究了移动机器人视觉定位的理论依据。首先介绍了摄像机成像与四种坐标系之间的关系。建立了相机成像模型,标定相机内参数。在此基础上,推导了移动机器人视觉定位技术相关的POSIT算法;第三,由于惯导和单目SLAM属于局部定位方法,缺少全局信息从而累积误差无法修正。本文研究融合惯导的视觉同时定位与地图构建方法,提出基于ORB_SLAM系统融合IMU信息的视觉惯导SLAM算法(VI_SLAM),并推导了视觉惯导SLAM系统的初始化算法。定位实验部分采用控制变量法,很好地验证了VI_SLAM融合算法的优越性和可靠性;第四,介绍整个平台的硬件与软件系统和实验部分。首先介绍了实验中使用的硬件和软件平台,并建立了双轮差速移动机器人的运动模型。简述了ROS操作系统及其常用功能,以及导航与定位的基础知识。实验的分析和设计考虑了多种不同情况的对比。提供了有效的,误差和定位精度的分析方法。实验分析表明,惯性导航和单目SLAM方法的结合有效提高了定位精度,展示了该算法的实际应用前景。
[Abstract]:The localization of mobile robot is the basis of autonomous navigation and an important research topic in the field of mobile robot. The main research content of this paper is based on ROS operating system, the inertial navigation and vision sensor information fusion. To solve the problem of autonomous positioning of mobile robot. At present, monocular SLAM has initialization scale problem and tracking scale drift problem, inertial navigation data is stable, But the accumulation error is more serious. The fusion of these two techniques can combine the high precision of monocular vision SLAM and the stability of inertial navigation data, learn from each other, and achieve the purpose of satisfying the positioning accuracy. The specific research contents are as follows: first, The basic positioning theory of strapdown inertial navigation system, including inertial navigation coordinate system, is studied. The basic theory of rotation, the knowledge of attitude solution based on quaternion algorithm and the theory of dead-reckoning lay the foundation for further research and application. The positioning accuracy of inertial navigation is tested experimentally. The theoretical basis of vision positioning of mobile robot is studied. Firstly, the relationship between camera imaging and four coordinate systems is introduced. The camera imaging model is established and the camera internal parameters are calibrated. The POSIT algorithm related to the vision positioning technology of mobile robot is deduced. Thirdly, because inertial navigation and monocular SLAM belong to local localization method, Because of the lack of global information, the accumulated error can not be corrected. In this paper, the visual simultaneous location and map construction method of fusion inertial navigation system are studied. A visual inertial navigation (SLAM) algorithm based on ORB_SLAM system fusion IMU information is proposed, and the initialization algorithm of visual inertial navigation (SLAM) system is derived. The control variable method is used in the positioning experiment, which verifies the superiority and reliability of VI_SLAM fusion algorithm. 4th, The hardware and software system and experiment part of the whole platform are introduced. Firstly, the hardware and software platform used in the experiment is introduced, and the motion model of the two-wheel differential mobile robot is established. The ROS operating system and its common functions are briefly described. And the basic knowledge of navigation and positioning. The analysis and design of the experiment take into account the comparison of many different situations. It provides an effective analysis method for error and positioning accuracy. The experimental analysis shows that, The combination of inertial navigation and monocular SLAM method can effectively improve the positioning accuracy and demonstrate the practical application prospect of the algorithm.
【学位授予单位】：哈尔滨工业大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.41;TP242
，

本文编号：1593582