服务机器人二维激光里程计构建及自主导航
发布时间：2019-04-20 11:56
【摘要】：近年来,服务机器人正逐渐走进千家万户,也不断改变着人们的日常生活方式。自主导航是服务机器人完成其它复杂任务的前提。本文研究了服务机器人在未知环境中基于二维激光测距的自主导航问题,其重点包括位姿跟踪、自主定位、栅格地图构建和路径规划。为有效实现移动机器人的位姿跟踪,对Mb-ICP、PSM口PL-ICP这三种典型的扫描匹配算法进行对比分析,并确定采用PL-ICP算法作为最终里程计构建算法。为减少原始杂乱数据对算法的影响,分别采用中值滤波和分割处理对激光数据进行预处理,剔除杂乱点和无效点,从而提高位姿跟踪算法的鲁棒性。此外利用关键帧技术解决由激光数据动态误差引起的位姿漂移问题。采用Rao-Blackwellized粒子滤波算法将机器人同时定位与地图构建(SLAM)问题分解为机器人全局位姿估计和栅格地图构建两个部分。为了改善粒子滤波算法的建议分布性能,采用激光里程计信息代替码盘里程计,从而在不影响算法精度的前提下减少了粒子的数量。此外根据粒子重要性权重的离散程度决定是否进行重采样,从而有效避免了粒子衰退现象。基于栅格地图各个栅格之间状态独立的假设,将栅格地图构建问题转化为估计每个栅格后验概率问题。考虑到机器人自身尺寸对实际导航的影响,根据机器人尺寸在栅格地图中将静态障碍物和实时检测到的动态障碍物进行膨胀处理。以代价地图、机器人位姿和目标点为基础,采用A*算法进行全局路径规划,给出从当前位姿到目标位姿的全局路径规划结果。在全局路径的指导下,采用动态窗口法进行局部路径规划,该方法可在速度空间中完成最优速度的搜索。此外引入机器人动力学约束,可进一步减小搜索空间,只保留可达速度,并最终基于评价函数选出最优局部路径解。在室内大范围环境中,以Pioneer DX3移动机器人为硬件平台,以ROS为基础构建服务机器人的自主导航软件系统。面向不同室内场景开展实验验证工作,试验结果表明本文所提方法的有效性和实用性。
[Abstract]:In recent years, service robots are gradually entering thousands of households, but also constantly changing people's daily life style. Autonomous navigation is the prerequisite for the service robot to complete other complex tasks. In this paper, the autonomous navigation problem of service robot based on two-dimensional laser ranging in unknown environment is studied. The emphasis includes pose tracking, autonomous positioning, grid map construction and path planning. In order to realize the pose tracking of mobile robot effectively, the three typical scanning matching algorithms, Mb-ICP,PSM port PL-ICP, are compared and analyzed, and the PL-ICP algorithm is chosen as the final odometer construction algorithm. In order to reduce the influence of the original clutter data on the algorithm, median filtering and segmentation are used to pre-process the laser data, and the clutter points and ineffective points are eliminated, so as to improve the robustness of the position-and-pose tracking algorithm. In addition, the key frame technique is used to solve the pose drift caused by the dynamic error of laser data. The Rao-Blackwellized particle filter algorithm is used to decompose the (SLAM) problem of robot simultaneous location and map construction into two parts: global pose estimation and grid map construction. In order to improve the proposed distribution performance of particle filtering algorithm, laser odometer information is used instead of disk odometer to reduce the number of particles without affecting the accuracy of the algorithm. In addition, resampling is determined according to the discrete degree of particle importance weight, so that the particle decay can be avoided effectively. Based on the assumption that each grid is independent of each grid, the problem of constructing raster map is transformed into the problem of estimating the posterior probability of each grid. Considering the influence of robot size on actual navigation, static obstacle and real-time detected dynamic obstacle are expanded in grid map according to robot size. Based on the cost map, robot pose and target point, the global path planning is carried out by using the A * algorithm, and the global path planning results from the current position to the target position are given. Under the guidance of the global path, the dynamic window method is used to carry out the local path planning. This method can search the optimal speed in the velocity space. In addition, the search space can be further reduced by introducing robot dynamics constraints, and only the reachable speed can be retained. Finally, the optimal local path solution can be selected based on the evaluation function. In the large-scale indoor environment, the autonomous navigation software system of service robot is constructed based on ROS and Pioneer DX3 mobile robot as hardware platform. Experiments are carried out on different indoor scenes, and the experimental results show that the method proposed in this paper is effective and practical.
【学位授予单位】：大连理工大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：TP242
本文编号：2461588
[Abstract]:In recent years, service robots are gradually entering thousands of households, but also constantly changing people's daily life style. Autonomous navigation is the prerequisite for the service robot to complete other complex tasks. In this paper, the autonomous navigation problem of service robot based on two-dimensional laser ranging in unknown environment is studied. The emphasis includes pose tracking, autonomous positioning, grid map construction and path planning. In order to realize the pose tracking of mobile robot effectively, the three typical scanning matching algorithms, Mb-ICP,PSM port PL-ICP, are compared and analyzed, and the PL-ICP algorithm is chosen as the final odometer construction algorithm. In order to reduce the influence of the original clutter data on the algorithm, median filtering and segmentation are used to pre-process the laser data, and the clutter points and ineffective points are eliminated, so as to improve the robustness of the position-and-pose tracking algorithm. In addition, the key frame technique is used to solve the pose drift caused by the dynamic error of laser data. The Rao-Blackwellized particle filter algorithm is used to decompose the (SLAM) problem of robot simultaneous location and map construction into two parts: global pose estimation and grid map construction. In order to improve the proposed distribution performance of particle filtering algorithm, laser odometer information is used instead of disk odometer to reduce the number of particles without affecting the accuracy of the algorithm. In addition, resampling is determined according to the discrete degree of particle importance weight, so that the particle decay can be avoided effectively. Based on the assumption that each grid is independent of each grid, the problem of constructing raster map is transformed into the problem of estimating the posterior probability of each grid. Considering the influence of robot size on actual navigation, static obstacle and real-time detected dynamic obstacle are expanded in grid map according to robot size. Based on the cost map, robot pose and target point, the global path planning is carried out by using the A * algorithm, and the global path planning results from the current position to the target position are given. Under the guidance of the global path, the dynamic window method is used to carry out the local path planning. This method can search the optimal speed in the velocity space. In addition, the search space can be further reduced by introducing robot dynamics constraints, and only the reachable speed can be retained. Finally, the optimal local path solution can be selected based on the evaluation function. In the large-scale indoor environment, the autonomous navigation software system of service robot is constructed based on ROS and Pioneer DX3 mobile robot as hardware platform. Experiments are carried out on different indoor scenes, and the experimental results show that the method proposed in this paper is effective and practical.
【学位授予单位】：大连理工大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：TP242
【参考文献】
相关期刊论文 前4条
1 刘兴权;金美含;;地图匹配算法综述[J];科技信息;2014年04期
2 孙志杰;王善军;张雪鑫;;工业机器人发展现状与趋势[J];吉林工程技术师范学院学报;2011年07期
3 徐景硕;周胜明;蒋华君;;惯性导航系统的发展及其关键技术综述[J];科技信息;2009年35期
4 关桂霞;朱虹;关永;吴敏华;;GPS定位误差分析与状态估计[J];计算机工程;2008年01期
，本文编号：2461588
本文链接：https://www.wllwen.com/kejilunwen/zidonghuakongzhilunwen/2461588.html
