CASIA OpenIR  > 毕业生  > 硕士学位论文
多机器人系统任务分配与视觉感知研究
任亮
学位类型工程硕士
导师谭民
2018-05-18
学位授予单位中国科学院研究生院
学位授予地点北京
关键词多机器人系统 任务分配 分层式组织结构 视觉感知 识别与跟踪
摘要多机器人系统通过群体协调可以完成单机器人无法完成的复杂任务,具有重要的研究价值和广阔的应用前景。任务分配是多机器人系统高效协作的基础,而多机器人系统对各机器人获取视觉信息的有效融合也将提升协同作业的质量。本文针对多机器人系统任务分配与视觉感知开展研究,主要内容如下:
首先,阐述了多机器人系统任务分配与视觉感知的选题背景和研究意义,对代表性的多机器人系统进行介绍,综述了多机器人系统任务分配与视觉感知的研究现状,并对论文的内容结构与章节安排做了说明。
其次,提出了一种基于资源模型与分层式组织结构的任务分配方法。该方法建立了资源向量的整合、比较和匹配规则,并构建了由底层机器人和上层管理者组成的分层式组织结构。在此基础上,根据任务约束条件进行机器人筛选和自底向上的管理者资源整合更新,并按照自顶向下的顺序将任务需求与管理者拥有的资源进行比较匹配以产生候选机器人联盟,最终选拔最优匹配度的机器人联盟执行任务。该方法能够有效处理复杂任务与大规模机器人系统的任务分配问题,提高了任务分配效率,具有良好的资源优化能力,并通过仿真进行了验证。
第三,在基于特征融合的单机器人目标识别方法与基于尺度自适应的单机器人目标跟踪方法基础上,提出了一种基于投票决策与权重自适应更新的多机器人系统视觉感知方法。该方法建立了多机器人系统视觉感知投票决策机制,并根据跟踪置信度与特征测度占比分别对各机器人的投票权重和特征的测度权重进行自适应更新,使具有较高准确性的机器人及特征来主导视觉感知过程。实验结果表明该方法实现了多机器人系统在目标识别与跟踪中的协同配合,提高了多机器人系统视觉感知的准确性。
最后,对本文工作进行了总结,并指出了需要进一步开展的研究工作。
其他摘要Multi-robot systems can accomplish complex tasks that cannot be completed by a single robot through the coordination among the robots, which is significant in both research and applications. Task assignment is the basis of cooperation among multiple robots, and the multi-robot system also needs to effectively integrate the visual information captured by each robot for the improvemnent of coordination. This thesis focuses on the task assignment and visual perception for multi-robot system. The contents are as follows:
Firstly, the research background and its significance of the multi-robot system task allocation and visual perception are given. Some typical examples of the multi-robot system are presented, and the research development of the task allocation and visual perception is then reviewed. The contents and structure of this thesis are also introduced.
Secondly, the task allocation method based on resource model and hierarchical framework is proposed. The proposed method establishes the rules for integration, comparison and matching of resources vectors. The hierarchical framework is built, which is composed of individual robots in the bottom layer and managers in higher layers. On this basis, the selection for robots as well as the bottom-up resources updating for each manager is executed according to the task resource constraint. Then the top-down resources comparison process between the task and the managers is used to generate candidate robot coalitions. Finally, the robot coalition with the optimal matching degree is selected to perform the task. The proposed approach can effectively deal with the challenges from complex tasks and large-scale multi-robot systems, and achieves a fast task allocation with optimal resources deployment, which are verified by simulations.
Thirdly, based on the target recognition method with feature fusion and the target tracking method with scale-adaptive for single robot, the visual perception method based on voting decision and adaptive updating of weights is proposed for multi-robot system. The voting decision mechanism of visual perception is made for multi-robot system, where for each robot, the voting weight and the feature measuring weight are updated according to the tracking confidence level and the measuring proportion, respectively. This solution can lead to a fact that the robot and feature with high accuracy dominate the visual perception process. The experimental results show that the proposed approach can realize the coordination of multi-robot system in target recognition and tracking with an improved accuracy of visual perception for multi-robot system.
Finally, the conclusions are given and future work is addressed.
文献类型学位论文
条目标识符http://ir.ia.ac.cn/handle/173211/20937
专题毕业生_硕士学位论文
作者单位中国科学院自动化研究所
推荐引用方式
GB/T 7714
任亮. 多机器人系统任务分配与视觉感知研究[D]. 北京. 中国科学院研究生院,2018.
条目包含的文件
文件名称/大小 文献类型 版本类型 开放类型 使用许可
任亮学位论文.pdf(3846KB)学位论文 暂不开放CC BY-NC-SA请求全文
个性服务
推荐该条目
保存到收藏夹
查看访问统计
导出为Endnote文件
谷歌学术
谷歌学术中相似的文章
[任亮]的文章
百度学术
百度学术中相似的文章
[任亮]的文章
必应学术
必应学术中相似的文章
[任亮]的文章
相关权益政策
暂无数据
收藏/分享
所有评论 (0)
暂无评论
 

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。