多主体环境MAGE及其协作与规划研究

发布时间：2018-05-16 08:39

本文选题：多主体系统 + 规划　；参考：《中国科学院研究生院（计算技术研究所）》2000年博士论文

【摘要】：多主体系统是近年来人工智能研究的热点，由于它更能体现人类的社会智能，更加适合开放、动态的世界环境，因而引起人们的广泛重视。在同一时期Internet技术的迅猛发展和普及应用使得软件开发人员迫切需要开放的、复杂的分布式计算系统的设计指导和开发辅助工具。在这两种力量的推动下，面向主体的程序设计也成为一个新的热点前沿课题。主体社会行为—特别是协作行为—的描述和实现则是其中一个重要研究方向。本文以主体之间的协作与多主体规划为对象，研究了主体协作的形式化描述，和基于约束传播的多主体分布式规划，提出了一种面向主体协作的组织形式，构建了多主体系统开发平台MAGE，并研究了主体技术的实用问题。本文成果主要包括如下几个方面： 1．提出了一种描述主体协作的形式系统。形式化工作被分为两个层次：底层是基本的逻辑体系，刻画了动作，知识，能力，目标等基本概念。在其基础上，定义了协作、多主体系统的潜在能力、协作求解框架等高层概念。 2．研究了基于约束传播的分布式多主体规划。扩展了ADL使之可以描述主体之间的联合动作；给出了动作之间的偏序约束的一致性判断方法：设计了分布式规划机制并给出一种算法实现。 3．提出了协作主体团队这一协作组织形式，并应用于多主体环境MAGE中。协作主体团队按照协作网络动态组成。意图协作的主体通过克隆操作加入。提供了协作团队形成、维护的标准协议和行为，并支持协作过程的并发执行。 4．针对目前对多主体系统开发工具的需求，设计并实现了主体系统环境MAGE。它为Internet上的基于主体的计算提供了系统设计支持，知识和能力的描述和装配，以及协作和协商的设计和实现提供了接口和工具。系统提供的名字服务，安全机制，查询和广告服务等标准服务方便了对多主体系统的管理和维护。 5．利用本文的研究成果，，特别是多主体环境MAGE，初步实现了并行多媒体信息检索、群体智能决策系统、虚拟市场等若干应用系统。综上所述，本论文对多主体协作与规划，多主体系统开发环境做了一定的研究。这些研究课题还有很多问题值得探讨，本文的工作仅仅是一个尝试，还有待进一步深入。
[Abstract]:Multi-agent system is a hot topic of artificial intelligence research in recent years. Because it can better reflect the social intelligence of human beings and is more suitable for the open and dynamic world environment, people pay more attention to it. In the same period, with the rapid development and popularization of Internet technology, software developers are in urgent need of open and complex distributed computing system design guidance and development assistance tools. Driven by these two forces, agent-oriented programming has become a new hot topic. The description and implementation of social behavior, especially cooperative behavior, is one of the important research directions. In this paper, the formal description of agent collaboration and multi-agent distributed planning based on constraint propagation are studied, and an organization form of agent-oriented collaboration is proposed. A multi-agent system development platform (MAGEG) is constructed, and the practical problems of agent technology are studied. This paper mainly includes the following aspects: 1. A formal system for describing agent collaboration is proposed. Formal work is divided into two levels: the basic logic system at the bottom, which describes the basic concepts of action, knowledge, ability, goal and so on. On the basis of it, the high-level concepts such as cooperation, potential capability of multi-agent system, cooperative solution framework and so on are defined. 2. The distributed multi-agent programming based on constraint propagation is studied. ADL is extended to describe the joint action between agents, and the consistency judgment method of partial order constraints between actions is given. A distributed planning mechanism is designed and an algorithm is implemented. 3. The collaborative agent team is proposed and applied to MAGE. The collaboration subject team is composed dynamically according to the collaboration network. The subject that intends to cooperate is joined by cloning operation. Provides collaborative team formation, maintenance of standard protocols and behaviors, and supports concurrent execution of collaborative processes. 4. In order to meet the need of developing tools for multi-agent system, we design and implement MAGE-based system environment. It provides interface and tools for system design support, description and assembly of knowledge and capability, and design and implementation of collaboration and negotiation for agent-based computing on Internet. The standard services, such as name service, security mechanism, query and advertisement service, are convenient for the management and maintenance of multi-agent system. 5. Some application systems such as parallel multimedia information retrieval, swarm intelligence decision system, virtual market and so on are implemented using the research results of this paper, especially MAGE-based multi-agent environment. To sum up, this paper does some research on multi-agent collaboration and planning, multi-agent system development environment. There are still many problems to be discussed in this paper. The work of this paper is only an attempt, which needs to be further explored.
【学位授予单位】：中国科学院研究生院（计算技术研究所）
【学位级别】：博士
【学位授予年份】：2000
【分类号】：TP18

【引证文献】