【作者】 骆正虎；

【导师】 蒋建国；

【作者基本信息】 合肥工业大学 ， 计算机应用技术， 2002， 博士

【摘要】 移动Agent提供了一种全新的分布式计算范型。移动Agent技术给分布式系统的设计、实现和维护都带来了新的活力。本文侧重于对移动Agent所涉及到的最根本问题进行深入的探讨，总结归纳出移动Agent系统的三个关键问题：通信机制、路由规划和联盟策略，并把它们作为全文的主要研究内容，以期推动移动Agent和分布式计算模型理论和技术的发展。本文的主要成果及创新之处如下： (1)利用Java语言的先进特性构建了一个移动Agent开发平台MADP。在MADP系统结构中，底层由TCP/IP协议、操作系统和Java虚拟机组成：AgentHost层为整个系统的核心层，由Agent管理器、安全服务、通信服务、迁移服务、事件服务和资源管理等模块组成，利用底层提供的服务向运行在其中的各移动Agent提供安全、顺畅的运行支撑环境；最上层是Agent层，由用户开发的运行在AgentHost环境中的各移动Agent组成，它们利用AgentHost层所提供的各种服务实现用户赋予的分布式计算任务。MADP系统的开发完全遵照面向对象的思想，系统具有良好的开放性，可以在其中通过加入模块的方法来实现智能性、增加安全性等。整个系统高度稳定、灵活性强、二次开发简便。 (2)提出一种高效的移动Agent消息传递机制DCC，将整个网络划分成多个域，通过主机、移动Agent及通信中介服务器之间的合作来实现安全高效的通信机制，可以实现移动Agent间的自由通信。不论目标Agent处于何种状态——“静止状态”、“迁移状态”还是“离线状态”，都可接收到发送给它的消息，从而很好地解决了移动设备断续连接的问题。为了消除因集中点的存在而引发的瓶颈问题，引入Agent当前位置缓存的概念并定义了有关缓存信息的四种通知消息，这样不但减轻了通信中介服务器的负担，将原本高度集中于通信中介服务器的负荷向整个网络进行均衡，而且还减小了消息传递的传输延迟、提高了通信效率，确保了系统的稳定性和可伸缩性，真正地实现了一种分布式的通信机制。 (3)在蚁群系统的基础上，通过引入变异算子和修正全局更新规则，提出一种改进型蚁群算法，蚂蚁之间通过外激素进行间接交流从而达到合作的目的，在利用已有信息与探索新解并重的策略指导下给出所求解问题的最优解，并且由于遗传算子的引入及全局更新规则的修正，解的多样性和随机性得到提高，不再易于陷入局部极小，在很大程度上避免了蚁群系统进化到某一伪最优解就停滞不前的现象。本文采用改进型蚁群算法求解移动Agent系统中重要的路由规划问题——旅行Agent问题，获得了满意的效果，实验结果表明本算法具有鲁棒性强、自适应、并行化、正反馈的优点，从而使得移动Agent在执行分布式计算任务的过程中，能够以最优的效率和最短的时间来实现预定的目标。 摘 要 （）提出一种基于遗传算法的移动 Agent系统联盟机制。Agent联盟作为移动 Agent系统中一种重要的协调与合作手段，其工作机理已得到越来越多的重视。本文在充分考虑Agent联盟问题的二维本质特性的基础上，提出了一种与之相适应的全新二维二进制染色体编码方式以及相应的二维或交叉算子和变异算子。这种二维编码方式及遗传算于突破了传统一维遗传算法的限制，使得合法亲代个体的遗传信息可以有效地融合到子代之中，并且所产生出的子代个体都是有效个体。同时还引入孤岛模型以提高算法的并行性和降低陷入局部最优的可能性，实验结果表明以上措施的引入大大提高了计算效率，缩短了计算时间，整个算法具有鲁棒性强、自适应性好、收敛速度快的优点。更多还原

【Abstract】 Mobile agent provides a novel paradigm for distributed computing. It has the potential to offer a single, general framework in which a wide range of distributed systems can be implemented efficiently, easily and robustly. Nevertheless, mobile agent technology is in a fairly early state. To make mobile agents attractive in as wide a range of applications as possible, a number of problems need. This dissertation mainly focuses on these problems and addresses three key issues of mobile ’agent system: communication mechanism, itinerary planning and coalition strategy. The main contributions of this dissertation are as follows:(1) A Mobile Agent Development Platform (MADP) is constructed based on the advanced characteristics of Java. In the system architecture of MADP, the bottom layer is composed of TCP/IP protocol, operating system and Java virtual machine. The core layer of the system, AgentHost layer, consists of agent manager, security service, communication service, migration service, event service and resource management. AgentHost provides a secure and smooth supporting environment for mobile agents. The agent layer, which is made up of those mobile agents developed by users, performs distributed computing assigned by their owners. The development of MADP conforms to the object-oriented principle. This makes the MADP system open enough for importing modules to enhance intelligence and security. The system is highly robust and flexible. It is very easy for users to develop their own distributed computing task using MADP.(2) This dissertation proposes DCC, an efficient message delivery mechanism for mobile agent system. The DCC mechanism separates the network into several domains. Hosts, mobile agents and communication and mediation server cooperate with each other to implement the communication among agents. Whatever status the target agent is in, it could receive messages. So it solves the intermittent connection problem of mobile devices successfully. To remove the bottleneck caused by centralized control, we introduce the concept of agent current location cache and define four notification messages about cache information. It eases the burden of communication and mediation server and shortens the transmission delay of message delivery. This method improves the communication efficiency and guarantees the robustness and scalability of mobile agent system.(3) An improved ant colony algorithm is presented in this dissertation. A mutation operator is introduced and global updating rule is modified on the basis of ant colony system. Ants could achieve the goal of cooperation by indirect communication mediated by pheromone and find goodIIIsolutions to , their task guided by the tradeoff between exploitation and exploration. The probability of premature convergence is low due to the introduction of genetic operator and modification of global updating rule. The traveling agent problem is solved using improved ant colony algorithm and promising result is obtained. Experimental results show that improved ant colony algorithm exhibits some excellent characteristics of robustness, self-adaptation, parallelism and positive feedback. So mobile agents could accomplish their distributed computing task with high efficiency and short time. (4) A coalition mechanism for mobile agent system is presented based on genetic algorithm. As an important coordination and cooperation mechanism in mobile agent system, agent coalition exhibits some excellent characteristics and receives attention increasingly. An efficient algorithm is needed for this topic since the number of the possible coalitions is exponential. Genetic Algorithm has been widely reckoned as a useful tool for obtaining high quality or even optimal solutions for a broad range of combinatorial optimization problems due to its intelligent advantages of self-organization, self-adaptation and inherent parallelism. This dissertation proposes a GA-based algorithm with an objective to simultaneously meet the goals of high performance, scalability, and fast run更多还原