QQ客服 官方微博 反馈留言

题名	空地机器人协作环境建模与路径规划
作者	梅元刚
学位类别	硕士
答辩日期	2014-05-28
授予单位	中国科学院沈阳自动化研究所
导师	何玉庆
关键词	空地协作 点云配准 路径规划 环境建模
其他题名	Aerial Robot and Ground Robot Cooperate to Environment Modeling and Path Planning
学位专业	控制工程
中文摘要	当前，移动机器人系统在广域野外非结构环境下的自主行为能力已成为移动机器人研究的关键问题之一，主要包括两个核心问题：一是如何获得机器人所处的环境信息；二是如何针对特定机器人以实时环境为约束进行行为规划。相对于单体机器人有限的环境感知能力，空地机器人协作能够有效的提高机器人系统的自主行为能力，受到国内外研究人员的关注。但是由于空地机器人环境感知的视角和尺度差异大导致现有空地协作环境建模仍然存在效率低和鲁棒性差等问题，同时地面机器人局部路径规划中如何充分利用融合的环境信息仍然是一个难点。针对如何提高空地机器人协作环境建模能力，本文首先提出一种以低维特征空间为引导的旋转图像点云地图快速配准方法。该方法首先建立低维特征空间，并在低维特征空间中搜索候选对应点集合，然后再在候选对应点集合中利用高维的旋转图像特征来搜索对应点对，从而实现点云地图配准。该方法能够有效提高点云地图配准的计算效率和鲁棒性。实验证明了本文提出的以低维特征为引导的旋转图像点云配准方法比经典旋转图像方法和多分辨率旋转图像方法的实时性和鲁棒性都有有效改善。但是在实际应用中，基于点云特征（旋转图像）的配准方法仍然难以满足移动机器人实时行为优化的要求。为此，本文融合改进的旋转图像点云配准方法的高鲁棒性和正态分布变换点云配准方法的高效率的优点，提出一种快速有效的点云地图序列化配准定位方法。实验证明了该方法的有效性。空地机器人系统中如何有效利用具有不同特征的环境数据，并实现地面机器人的高效局部路径规划，是空地协作中的一个研究热点。针对空地环境地图具有不同的视野范围和分辨率的特点，本文基于机器人运动学约束的路径搜索方法，提出一种以低分辨率全局最优路径代价图为引导的局部路径规划方法。该方法能够有效利用空中低分辨率全局地图的路径规划信息来引导地面机器人实现具有全局最优性的局部避障策略。通过仿真实验，证明了该方法能够提高地面机器人路径规划的全局有效性和计算效率。
索取号	TP242/M44/2014
英文摘要	At present, the autonomous behavior ability of the robots in the outdoor unstructured environments has become one of the key points in the research of mobile robots, mainly includes two core issues: one is how to obtain the robots’ environment model; the second is how to make the best planning on the constraint of the environment. Relative to limited environmental awareness of the monomer robot, the cooperation of unmanned aerial and ground vehicles can effectively improve the environmental awareness ability and perform better in some difficult applications, thus it has attracted many researchers’ attention. However, the complexity of the unstructured environment information, different perspective and scale between the heterogeneous robots may cause low efficiency and bad robustness.To improve the environmental perception of the air/ground robotic systems, this paper proposes a fast spin image based point clouds registration algorithm, the new algorithm is based on a low-dimensional feature space. The main conception of the method is that through constructing the low-dimensional feature space, the correspondence searching procedure can be divided into two steps: firstly, select a very limited amount of point candidates in the low-dimensional feature space; then the high dimensional spin image feature is used to search for correspondences among the point candidates. Experiments show that the proposed new registration method has effectively been improved in contrast to the classical spin image methods and multi-resolution spin image method.However, in the practical application, the features (such as spin images) based point cloud registration method is still difficult to meet the real-time requirements of the mobile robotics system. Combined the robust spin image based point cloud registration algorithm with the fast normal distribution transformation based point cloud registration algorithm, this paper propose a new fast and effective point cloud map sequence location algorithm. Experimental results show the validation and effectiveness of this method.How to use environmental information effectively which are in different characters and make a best local planning path for the ground robot is one of the key issues of the air and ground robotic systems. The air global maps and the ground local maps have different vision field and resolution. This paper proposed a local path planning theory based on global planning path optimal costs map, which also capable of satisfying the constraints on motion. This new proposed algorithm can use the low-resolution global information for ground robot planning a global optimal path effectively. Simulation results show that this method can improve the global optimality and computational efficiency of ground robot path planning.
语种	中文
产权排序	1
页码	65页
分类号	TP242
内容类型	学位论文
源URL	[http://ir.sia.ac.cn/handle/173321/14838]
专题	沈阳自动化研究所_机器人学研究室
推荐引用方式 GB/T 7714	梅元刚. 空地机器人协作环境建模与路径规划[D]. 中国科学院沈阳自动化研究所. 2014.

个性服务

查看访问统计

相关权益政策

暂无数据

收藏/分享

除非特别说明，本系统中所有内容都受版权保护，并保留所有权利。