【作者】 张伟；

【导师】 王雪帆；

【作者基本信息】 华中科技大学 ， 电气工程， 2018， 硕士

【摘要】 随着变电站无人值守的推广,变电站巡检机器人得到更加广泛的应用。变电站巡检机器人作为一种自主导航的移动机器人,其首先需要解决的是自主定位与建图问题。3D激光雷达作为一种高频率、高精度、强鲁棒性的三维环境感知传感器,已成为移动机器人定位与建图的研究热点。本文以广东电网“变电站智能巡检机器人”项目为依托,设计了一套基于3D激光雷达的定位与建图系统,在缺少GPS(Global Positioning System)的情况下,实现了机器人的精确定位和环境点云地图构建功能,主要完成了以下工作:首先,研究了机器人的姿态表示方法和基于点云匹配的运动估计原理,实验对比了几种点云匹配算法的准确性和时效性,证明了基于PTL&P(Point to Line&Patch)的匹配算法的优越性。然后,研究了基于滤波和优化两种方法的数学模型,提出了一种基于3D激光雷达的图优化同步定位与地图构建方法,通过回环检测进行全局图优化求解,修正点云匹配的累积误差,得到最优的机器人位姿估计和准确一致的地图,为基于地图的定位提供了相对准确的先验环境数据。其次,提出了一种基于点云地图的多传感器组合定位方法,运用互补滤波算法融合里程计、惯导的预测结果和3D激光雷达的观测结果,解决了机器人在无GPS环境下的三维空间定位问题。最后,以广东电网自主搭建的配有3D激光雷达、里程计、惯导等多传感器的移动机器人作为实验平台,在ROS(Robot Operating System)环境下对本文所提出的算法进行了多种环境下的实验验证,结果表明本文提出的定位与建图方法定位精度高,鲁棒性强,适用于变电站环境。更多还原

【Abstract】 With the extension of unattended substation,nowadays,the substation inspection robot has been more widely used.As an auto-navigating mobile robot,the first problem needed to be solved for substation inspection robot is self-localization and mapping.As a 3D environment perception sensor with high frequency,high precision and strong robustness,3D lidar becomes the research hotspot in localization and mapping of mobile robots.Under the project ‘Intelligent Substation inspection Robot ’ of Guangdong Grid,a localization and mapping system based on 3D lidar has been designed in this thesis,which can realize precise positioning and building the point cloud map of environment under the situation without GPS.The main works of this thesis are as followes:First,the attitude representation and motion estimating principle based on point cloud matching are researched.Meanwhile,the accuracy and time consuming of several point cloud matching algorithms are compared by experiments.The superiority of the matching algorithm based on PTL&P is proved.Then,the mathematical models of two methods based on filtering and optimizationa are researched.A new method of graph-based slam based on 3D lidar is proposed,by global graph optimizing solution with loop detection and modifying the accumulative error of point cloud matching,the optimal map and robot pose estimation can be obtained,which provides accurate transcendent environmental data for localization based on map.Next,a method of multi-sensor integrated positioning based on point cloud map is proposed.It can solve the problem of 3D special localization for robot without GPS by complementary filter algorithm which combines the prediction results of odometer and inertial navigation system,as well as the observation results of 3D lidar.Finally,based on the experimental platform of the mobile robot with 3D lidar,odometer,inertial navigation system and other multi-sensor that constructed by Guangdong Grid,the algorithms proposed in the thesis is tested under multiple environments.It shows that the proposed method of localization and mapping has high accuracy and strong robustness,which is suitable for the environment of substation.更多还原