CASIA OpenIR  > 数字内容技术与服务研究中心  > 听觉模型与认知计算
移动机器人导航控制系统应用研究
卫浩
Subtype硕士
Thesis Advisor李学恩
2019-05-23
Degree Grantor中国科学院大学
Place of Conferral中国科学院大学
Degree Name工程硕士
Degree Discipline工学硕士
Keyword玻璃检测算法 多传感器融合 机器人导航 激光雷达slam 视觉slam
Abstract

       随着技术的进步,机器人逐渐从传统工业生产环节进入到公共服务、家庭娱乐等场景中。导航控制技术是移动机器人的核心技术,是机器人完成复杂服务的基础。本文主要研究机器人的导航控制系统,针对现有激光雷达SLAM(Simultaneous Localization And Mapping)算法和视觉SLAM算法场景适应性低和鲁棒性差的问题,采用多传感器融合方法,提高了现有机器人导航系统的环境适应性和鲁棒性。本文的主要工作包括:
       第一,本文对现有激光雷达和视觉SLAM算法进行了实验和总结。本文构建了一种简易的激光SLAM算法,并对算法的效果进行了实验,结果表明,算法能够有效建立环境地图。本文从建图效果和计算资源消耗两个角度对比了三种激光SLAM算法的性能,实验表明,三种算法中Cartographer算法建图效果最好,但计算资源消耗也最高,Gmapping算法计算资源消耗较低,建图效果较好。本文也构建了一种基于视觉SLAM算法,视觉前端基于特征点方法,采用PnP(Perspective-n-Point)算法计算初始位姿,后端采用位姿图优化算法,回环检测基于近距离回环和全局随机回环,保证了回环检测的效率,最终构建出环境的点云地图。
       第二,为了解决低成本激光雷达无法有效检测玻璃的问题,本文提出了一种融合超声波传感器和激光雷达信息的玻璃检测算法,并对算法进行了实验。实验表明,机器人能够有效检测出大部分的玻璃,实验结果证明了算法的有效性。机器人采用带玻璃检测的激光SLAM算法的建图结果会降低定位准确度,本文采用路径规划和定位地图分离的方法,在保证机器人定位性能的同时,提高了机器人的导航效率。实验结果表明,玻璃环境下采用新的导航框架机器人导航效率明显提升。本文提出的玻璃环境下基于低成本激光雷达导航算法有实际应用价值。
       第三,针对现有视觉SLAM算法鲁棒性差的问题,本文提出了一种多传感器融合的视觉SLAM算法,算法将机器人的里程计和IMU(Inertial Measurement Unit)信息融合到视觉SLAM的图优化框架中,提高了纯视觉SLAM的鲁棒性。利用EKF(Extended Kalman Filter)算法融合机器人单轴陀螺仪和里程计数据,计算得到在两视觉关键帧之间的机器人坐标系位姿,将机器人坐标系位姿转换到相机坐标系,机器人坐标系位姿作为相机测量位姿的约束,对整个系统进行优化。本文对融合里程计和IMU信息的视觉SLAM算法进行了实验,实验结果证明了算法的有效性。

Other Abstract

   With the rapid advancement of technology, robots have gradually entered public services and home entertainment from traditional industrial factory. Navigation and control technology are the core technology of mobile robots and the basis for robots to complete complex services. This paper mainly studies the navigation and control system of the robot. Due to the existing Lidar based SLAM algorithm and the visual based SLAM algorithm have low adaptability and poor robustness, the multi-sensor fusion method is adopted to improve the adaptability and robustness of the existing robot navigation system. The main work of this paper includes:
   First, this paper experiments and summarizes the existing lidar based and vision based SLAM algorithms. In this paper, a simple laser-based SLAM algorithm is constructed and the effect of the algorithm is tested in the simulation environment. The performance of three laser SLAM algorithms is compared from the perspective of mapping effects and computational resource consumption in this paper. Experiments show that the Cartographer algorithm has the best mapping results, but the computing resource consumption is also the highest. The Gmapping algorithm calculates the resource consumption is lower, but the mapping effect is better. This paper also constructs a visual based SLAM system, the visual frontend is based on the ORB feature, the PnP algorithm is used to calculate the initial pose, and the backend adopts the pose optimization algorithm. The loop detection is based on the close loop and the global random loop, which ensures the loop detection efficiency, and ultimately build a point cloud map of the environment.
   Secondly, in order to solve the problem that low-cost laser radar cannot effectively detect glass, a glass detection algorithm is proposed in this paper that fusion ultrasonic sensor and lidar information, and experiments on the algorithm. Experiments show that the robot can effectively detect mostly of the glass in the environment, and the experimental results prove the effectiveness of the algorithm. The robot using the lidar SLAM algorithm results with glass detection will reduce the positioning accuracy. This paper adopts the method of path planning and location map separation to improve the navigation efficiency of the robot while ensuring the positioning performance of the robot. The experimental results show that the navigation efficiency of the new navigation frame is improved significantly in the glass environment. The low-cost lidar navigation algorithm based on the proposed glass environment has practical application value.
   Thirdly, aiming at the problem of poor robustness of the visual based SLAM algorithm, this paper proposes a multi-sensor fusion visual SLAM algorithm, which integrates the robot's odometer and IMU information into the visual SLAM map optimization framework, and improves the pure visual SLAM robustness. Firstly, the EKF algorithm is used to fuse the data of the robot single-axis gyroscope and the odometer, and the pose matrix of the robot coordinate system between the two visual key frames is calculated. The pose matrix of the robot coordinate system is converted to the camera coordinate system, and the pose matrix of the robot coordinate system is used. The camera measures the pose constraints and optimizes the entire system. In this paper, the algorithm is tested and the experimental results demonstrate the effectiveness of the algorithm.

Subject Area计算机科学技术其他学科
MOST Discipline Catalogue工学::计算机科学与技术(可授工学、理学学位)
Pages78
Language中文
Document Type学位论文
Identifierhttp://ir.ia.ac.cn/handle/173211/23816
Collection数字内容技术与服务研究中心_听觉模型与认知计算
Recommended Citation
GB/T 7714
卫浩. 移动机器人导航控制系统应用研究[D]. 中国科学院大学. 中国科学院大学,2019.
Files in This Item:
File Name/Size DocType Version Access License
最终版_移动机器人导航控制系统应用研究.(15966KB)学位论文 开放获取CC BY-NC-SAApplication Full Text
Related Services
Recommend this item
Bookmark
Usage statistics
Export to Endnote
Google Scholar
Similar articles in Google Scholar
[卫浩]'s Articles
Baidu academic
Similar articles in Baidu academic
[卫浩]'s Articles
Bing Scholar
Similar articles in Bing Scholar
[卫浩]'s Articles
Terms of Use
No data!
Social Bookmark/Share
All comments (0)
No comment.
 

Items in the repository are protected by copyright, with all rights reserved, unless otherwise indicated.