摘 要

随着近年来智能机器人技术的发展，扫地车作为服务型机器人的代表走进了千家万户，扫地机器人的环境感知和路径规划也成为了服务型机器人领域所研究的前沿和热点，也是实现智能扫地机器人的关键技术。目前主流的扫地机器人大多采用激光雷达作为主传感器进行环境感知，成本高昂，测量精度高但容易受到天气和光线的干扰，扫描距离不足，并且在全覆盖路径规划算法的选择上也存在一定不足，导致扫地机器人清扫的重复率高，覆盖率低。在此类机器人的开发中，各个厂家针对自己的元器件和产品进行开发优化，通用性不足，对个人、团队开发带来了一定的困扰。

针对上述问题，本文设计了基于视觉导航的无人扫地车，本文主要研究工作如下：

（1）在ROS机器人操作系统上完成系统框架的搭建，针对激光雷达成本高、距离短的问题，采用了双目视觉惯性模块进行扫地车的定位建图，能够获得范围更大、实用性更强、数据更全面的深度信息，使扫地车的环境感知范围更远，定位精度更高，成本更低。

（2）设计了基于单线激光雷达的避障模块，使用最小二乘法拟合障碍物半径，计算障碍物的长度、世界坐标等信息，能够精确地识别前方障碍物的类型，根据不同障碍物的特征采取不同避障策略。使用往复式全覆盖路径规划实现子区域的覆盖，并结合A*算法进行路径查找和死区逃离，优化扫地车的清扫效果。设计了基于CH340USB转串口芯片的上下位机通讯程序，实现扫地车控制指令的发送。

（3）在ROS可视化仿真环境中进行了建图和清扫效果的仿真，在两种不同的环境地图上进行测试，得到扫地车的综合覆盖率均在80%以上，满足预期设计效果。

关键词：ROS机器人；视觉SLAM建图；全覆盖路径规划

Abstract

With the improvement of the robot technology, the service robot represented by the cleaning robot has entered thousands of families, and the environmental perception and path planning of cleaning robot have become the forefront and hotspot of the research in the development of the service robot, as well as the key to realize the intelligent cleaning robot. At present, most of the cleaning robots use laser lidar as the main sensors for environmental perception, which is costly and has high measurement accuracy, but is easily interfered by weather and light. Besides, the scanning distance is insufficient, and the selection of the path planning algorithm for complete coverage is also inadequate, resulting in high repetition rate and low coverage rate of cleaning robots. In the development of such robots, each manufacturer develops and optimizes its own components and products, which brings some trouble to the development of individuals and teams.

To solve the problems above, this paper designed an unmanned cleaning robot based on vision navigation, The main research contents of this paper are as follows:

(1) using “ROS” robot operating system to complete the construction of the system framework. Aiming at the problem of high cost and short distance of lidar, the stereo camera and inertia module is used to build the positioning map of cleaning robot, which can obtain more deep information with larger range and more comprehensive data, so that the environmental perception range of cleaning robot is further, the positioning accuracy is higher and the cost is lower.

(2) An obstacle avoidance module based on single-line lidar is designed, which uses the least square method to fit the radius of obstacles, calculate the length and world coordinates of obstacles, so that the cleaning robot can accurately identify the types of obstacles and adopt different obstacle avoidance strategies according to the characteristics of different obstacles. The reciprocating complete coverage path planning is used to realize the coverage of cleaning areas, and the A* algorithm is combined to search the path and escape the dead zone, to optimize the cleaning effect of the cleaning robot. The communication program of the upper and lower computer based on CH340 USB to serial port chip is designed.

(3) The map building and cleaning effect simulation were carried out in the ROS visual simulation environment, and the test was conducted on three different environmental maps. The result shows that the comprehensive coverage of the cleaning robot was all above 80%, which meets the expected design effect.

Key Words：； Complete Coverage Path Planning; Visual SLAM; Navigation

目录

第1章 绪论 1

1.1 研究背景及意义 1

1.2 国内外研究现状 2

1.3 文章的研究内容与结构安排 5

第2章 基于双目摄像头的SLAM建图 7

2.1 扫地车的模型及坐标变换 7

2.2 基于视觉里程计和IMU惯性模组的融合定位 10

2.3 基于双目摄像头的三维点云建图 12

2.4 本章小结 15

第3章 扫地车避障算法的设计 16

3.1 障碍物检测方案设计 16

3.2 障碍物规避方案 22

3.3 本章小结 24

第4章 扫地车路径规划部分的设计 25

4.1 工作空间地图的栅格化处理 25

4.2 扫地车全覆盖路径规划算法的设计 27

4.3 算法仿真测试 30

4.4 本章小结 33

第5章 扫地机器人总体方案设计 34

5.1 基于ROS的机器人系统设计 34

5.2 基于ROS的扫地车系统设计 35

5.3 本章小结 40

第6章 总结与展望 41

6.1 全文总结 41

6.2 课题展望 41

参考文献 42

致谢 44

第1章 绪论