基于激光雷达的即时定位与制图技术

发布时间：2018-06-16 11:56

本文选题：机器人底盘 + 即时定位与制图　；参考：《哈尔滨工业大学》2017年硕士论文

【摘要】：随着科学技术的发展,人们生活水平的提高,社会各界对于机器取代人有着越来越高的期望,希望能有更加智能的设备出现,将人们从繁重的重复工作中解脱出来。然而早期的机器人功能单一,只能按照预先定义的程序工作,以固定的路线运动。真正的自主机器人要求具备的一项基本功能是可以识别自身在环境中的位置,这就要求机器人可以描绘出所在环境的地图模型,并确定自身在地图上的位置,然后才可以执行后续其他任务。而这种自主制图的功能就是即时定位与制图(SLAM)技术所要完成的任务。本文研究分析基于激光雷达的室内SLAM技术,并根据其框架建立了系统的概率模型、观测模型以及地图模型。针对室内环境相对简单的结构特征,以及激光雷达的数据特点,综合了现有的线段特征提取方法,提出一种新型线段特征提取方法,从激光雷达数据中提取线段特征,并用于构建增量式地图。本文对主要的滤波算法进行比较分析和仿真实验,研究这些算法对于状态估计的效果,然后对基于扩展卡尔曼滤波方法的SLAM算法进行仿真实验。最后用本文的线段特征提取方法融合进基于粒子滤波的SLAM算法,并对其权重更新规则和重采样方法进行优化。再通过仿真程序检验本文SLAM算法。针对室内环境较狭小的特点,研发设计能在室内灵活运动,原地转向的三轮全向机器人平台。使用机器人操作系统(ROS)作为机器人的软件开发平台,分析本文机器人系统的硬件框架以及模块间的通信关系,利用ROS强大的功能提高软件开发的效率,使用它的TF软件包高效管理机器人系统复杂的坐标系统。最后基于该机器人SLAM平台,在办公区的走廊上进行SLAM测试,实验结果验证了整个SLAM系统的高效性。
[Abstract]:With the development of science and technology and the improvement of people's living standard, people from all walks of life have higher and higher expectations for machines to replace people. The early robots, however, had a single function and could only work according to a predefined program, moving along a fixed route. One of the basic functions that a truly autonomous robot requires is to recognize its position in the environment, which requires that the robot be able to map its environment and determine its position on the map. Then you can perform other follow-up tasks. The function of this kind of independent mapping is the task to be accomplished by the technology of instant positioning and mapping. In this paper, the indoor slam technology based on lidar is studied and analyzed, and the probabilistic model, observation model and map model of the system are established according to its frame. Aiming at the relatively simple structural features of indoor environment and the characteristics of lidar data, this paper synthesizes the existing line segment feature extraction methods, and proposes a new line segment feature extraction method, which extracts the line segment feature from the lidar data. And used to build incremental maps. In this paper, the main filtering algorithms are compared and simulated, and the effect of these algorithms on state estimation is studied. Then, the slam algorithm based on extended Kalman filter is simulated. Finally, the line segment feature extraction method is integrated into the slam algorithm based on particle filter, and its weight updating rules and resampling method are optimized. Then the slam algorithm is verified by simulation program. Aiming at the narrow indoor environment, a three-wheeled omni-directional robot platform with flexible indoor movement and in-situ steering was developed. This paper analyzes the hardware frame of the robot system and the communication relationship between modules, and improves the efficiency of software development by using the powerful function of Ros. The TF software package is used to efficiently manage the complex coordinate system of the robot system. Finally, based on the slam platform of the robot, the slam test is carried out in the corridor of the office area. The experimental results verify the efficiency of the slam system.
【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】：TN958.98;TP242

【参考文献】