SIA OpenIR  > 空间自动化技术研究室
面向任务的群机器人系统分层自组织规划研究
其他题名Research on Hierarchical Self-Organization Planning of Task-Oriented Swarm Robotic System
冷雨泉1,2
导师周维佳
分类号TP242
关键词面向任务 群机器人 分层自组织 任务分解 任务分配
索取号TP242/L26/2016
页数125页
学位专业模式识别与智能系统
学位名称博士
2016-12-05
学位授予单位中国科学院沈阳自动化研究所
学位授予地点沈阳
作者部门空间自动化技术研究室
摘要纵观机器人发展趋势,可归纳出以下四方面趋势特点。a) 单机传统性能层面。电子信息技术及先进加工制造技术的发展,使机器人在精度、实时性、负载自重比等性能层面得到了提升;b)智能层面。人工智能技术的发展,促进了机器人解决问题能力的提高;c) 数量层面。机器人发展经历了从单机作业到多机协同作业,再到群机器人任务分配协作的趋势;d) 操作层面。传统的人工遥操作、人工试教的方式正逐步的被机器人自主操作所取代。本文的研究将涉及到后三方面的特点,针对群机器人在任务下达和解决方法进行研究,建立群机器人系统架构体系研究,以及提出面向任务的智能化解决方法研究。本文将围绕群机器人系统体系架构、任务下达方式、任务分解与分配、群自组织规划等方面进行研究。具体阐述如下:(1)群机器人概念最早出现于上世纪90年代,目前,对于群机器人的概念具有较多的定义方式,总结现有学者对群机器人概念的定义,并结合现阶段机器人技术的发展趋势,文中给出了更为广义的群机器人概念的定义。概念中,首先,对机器人的概念进行抽象化,具有一定的智能性的计算单元亦可以定义为机器人,雷同于Agent技术;另一方面,概念的核心是群机器人的组织及组织关系,不同于以数量为核心的传统概念。(2)系统的体系结构是一个系统的框架,文中提出群机器人系统在系统层面的三层控制体系结构;详细的阐述群机器人系统中的组织关系及个体在群系统中的社会抽象 ;描述了面向任务的群机器人系统中个体层的控制体系结构,明确个体任务执行的机理及决策的产生。(3)群机器人的优势在于依靠群体解决任务。命令下达者需要对任务进行确切的下达,针对现有的人机交互方式,对各类交互方式,从交互的舒适性、信息传递的准确性、交互的快速性、交互任务的复杂性等方面进行比较,提出图形化虚拟世界的任务交互方式;对于新的交互方式采用本体论理论对图形化虚拟世界中的模型进行形式化定义;基于本体论的模型并结合本体的代数系统,建立具体世界的树形表示;描述图形化虚拟世界任务下达方法的实现流程。(4)群机器人的面向任务性是群机器人系统必备的重要性能。基于一阶逻辑提出任务的概念化及符号表示,依靠一阶逻辑中的逻辑连接词,建立逻辑分解关系;提出任务间的时序约束关系建立时序网络;阐述了任务的分解方法及分解原则并基于Petri网建立任务分解的树形结构;提出机器人能力及任务需求的指标规范,建立任务与机器人之间任务实现性桥梁,便于群机器人的规划;提出基于代价函数的合作式的任务分配方法,作为文中提出的群机器人系统采用的通用的任务分配方式。(5)描述分层自组织模型的结构及特征,并确定组织中个体的定义方式;阐述了分层自组织方式的形成方法和形成规则,并利用基于改进的荷尔蒙数学模型模拟群体的分层自组织形成过程;提出分层式自组织对组织形成过程的性能指标,用于指导分层式自组织方法的优化;提出风蚀算法,解决自组织问题中出现的纵向层数过深的问题;验证基于改进的荷尔蒙数学模型模拟群体的分层自组织方法的自修复功能,其能够使系统解决组织中个体的损坏问题。(6)搭建SociBuilder平台,模拟人类社会的建造房屋的机理。通过群机器人系统中的个体搬运和操作积木模块搭建由用户基于图形化虚拟世界下达的任务;利用文中的理论,在SociBuilder系统中实现任务分解、分配与执行,验证系统面向任务的自组织规划能力,尤其是在面对个体受损情况的组织恢复能力,实现群机器人系统由人类下达任务到系统完成任务的智能建造。
其他摘要Retrospect the development of robots, four trends can be summarized from different view point: a) Individual robot performance level. Development of Electronic information technology and advanced manufacturing technology make robots more accurate, respond more quickly in real time and in parallel realize a low load weight ratio. b) Intelligence level. Development of artificial intelligence technology promotes robotic ability in solving problems. c) Quantity. Robot development has experienced from single operation to multi machine cooperation, and then to swarm robot task allocation and cooperation. d) Operation. Traditional teleoperation and artificial trial teaching are gradually being replaced by robot autonomous operation. In my research, the last three aspects will be deeply studied. The main purpose of this paper is trying to solving task assignment and task automatic execution of swarm robotic. System architecture, task assignment, task decomposition and allocation, self-organization planning of swarm robotic and so on will be discussed independently. Detailed descriptions are as follows: Conception of swarm robot first appeared in 1990s. Since then many definitions of concept of swarm robotic are put out. Summing the previous definitions and combined with the development trend of robot technology, the paper gives one more general definition of the swarm robot. First of all, the concept of robotic is abstracted, such as that a calculation intelligence unit can also be defined as robot, similar to Agent technology. On the other hand, the core of the proposed concept is organization and organization relationship, different from the traditional concept which always focus on the quantity. As for the framework of swarm robot, a three layer hierarchical control structure is put forward in this paper in detail. Organization relationship and individual abstraction in swarm robot system are described. Describing structure of individual layer for task oriented swarm robot system and mechanism for decision making and execution. Learning and planning mechanism of swarm robot control system are descripted. Task is solved by many swarm individual, which is the most advantage of swarm robotic. Human need to give accurate task information to system. Compare all kinds of human-computer interaction, including comfort, rapidity, accuracy, and so on. One suitable method for human-computer interaction in swarm system is proposed based on the task interaction in a graphical virtual world. The formal definition of the model in the graphical virtual world is made by using the theory of ontology. Establishing the tree representation of the concrete world based on ontology and algebraic system. Task oriented mechanism is a necessary and important property for swarm robot system. Based on the first order logic, concept of task and symbolic representation are proposed, and based on logical connective, logic decomposition relation is established. Task sequence network is set up for establishing relationship between the subtasks. According to Petri Net, task is decomposed to a tree structure. To map the task requirements to the robot’s abilities, metrics of robot abilities of task requirements are formulated, which is useful for task planning in latter stage. Cooperative task allocation method based on cost function is used as a general task allocation mehod in our paper. Structure and characteristics of hierarchical self-organization model are described, and the definition of the individual in the organization is proposed. Formation methods and rules of hierarchical self-organizing method are described, and the hierarchical self-organization forming process based on the improved mathematical model of hormone is also discussed. This paper puts forward the performance index of the hierarchical self-organization to the organization formation process, which is used to guide optimization of the hierarchical self-organizing method. Considering sometimes the number of the vertical layers is too big in self-organizing process, wind erosion algorithm is used to simplify the structure by cutting some extra layers. At last, this paper puts forward a self-repair algorithm for hierarchical self-organizing, so that the system can work normally in exceptional cases, for example, some individuals are damaged. A SociBuilder platform is built to simulate mechanism of human society building. The swarm robot system consists of two kind of individuals, one is used for transporting and the other one is used for manipulate blocks. The task is dispatched in a graphical virtual world which is very intuitive. Task decomposition, distribution and execution are implemented in the SociBuilder system to verify the self-organization capability of the system, especially when some individual damage happens. Finally, combining all these modules together, the blocks building task is fulfilled smartly and adaptively form the very beginning to the end.
语种中文
产权排序1
文献类型学位论文
条目标识符http://ir.sia.cn/handle/173321/19452
专题空间自动化技术研究室
作者单位1.中国科学院沈阳自动化研究所
2.中国科学院大学
推荐引用方式
GB/T 7714
冷雨泉. 面向任务的群机器人系统分层自组织规划研究[D]. 沈阳. 中国科学院沈阳自动化研究所,2016.
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
面向任务的群机器人系统分层自组织规划研究(4092KB)学位论文 开放获取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
[冷雨泉]的文章
百度学术
百度学术中相似的文章
[冷雨泉]的文章
必应学术
必应学术中相似的文章
[冷雨泉]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。