智能虚拟环境是人工智能和虚拟现实相结合的产物。由于多Agent系统具有可扩展性、自治性、主动性、可重构性和敏捷性等特点,已成为目前最感兴趣的AI技术之一。本文提出一种IVE中基于多Agent系统的智能Agent体系结构,着重于智能Agent内部机制的研究与设计,通过Agent的内部机制来提升系统的整体性能。

Intelligent Virtual Environment is an outcome of Artificial Intelligence and Virtual Environment. Due to its features:scalable, autonomous, proactive, reconfigurable, agile, the Multi-Agent Systems (MAS) has been the most interesting AI technique. This paper presents a new Agent architecture for Intelligent Virtual Environment base on MAS, concentrates on the internal mecha-nism research and design of intelligent Agent, and promotes the whole performance of the system by Agent’s internal mechanism.

针对多机器人远程监控系统信息错综复杂、协作不稳定的问题,建立基于多智能体(Multi-Agent)技术的系统体系结构,描述系统各组件之间的信息交互关系,优化人-机智能分配。分析系统物理结构的特点,提出一种基于Multi-Agent技术的共享控制系统分层体系结构,结合黑板结构和点对点结构,给出Multi-Agent的混合通信模型。针对遥操作系统的特点,设计一种混合型Agent体系结构,举例研究Agent的实现方法。通过多操作者控制机器人保持队形的实验,验证了该混合型Agent体系结构的实用性和有效性。

According to the problem of multi-robot intricate and multi-robot collaboration unstable for remote monitoring system of multi-mobile robot, a system architecture based on multi-Agent technology can be established to optimize the man-machine intelligence allocation and to give a clear description of the various components of the system. By the analysis of the physical structure of the system, the architecture of hierarchical system based on share control mode of multi-Agent is proposed and a mixed communication model of multi-Agent combining the blackboard architecture and peer-to-peer structure is designed. According to the characteristics of the teleoperation system, a hybrid Agent architecture is designed. An example of the realization of the Agent is given, and an experiment that multi-operator keep robots controlled in line is designed to verify the effectiveness of the hybrid Agent architecture.

将多Agent理论和技术应用于联合灭火救援作战的研究,构建了多Agent联合灭火救援作战模型框架,实现了联合灭火救援组织指挥相互之间的信息交换,显示了互相影响的行动过程.对系统中指挥Agent、作战Agent和环境Agent进行描述,定义Agent之间以及Agent与环境之间的交互规则,来描述系统的相互关系,给出了多Agent的协同算法,并构建了联合灭火救援仿真模型.最后用多Agent软件工具集和通用软件平台(SWARM),采用JAVA语言编程,进行了模拟仿真,对文中构造的模型及相应机制进行了功能验证.

A framework of multi-agent joint firefighting and rescue operation was proposed by applying the multi-agent theory and technology to studies on joint firefighting and rescue operation. It can realize information exchange between the commanders and display the interacted course of operation. Based on the description of command agent, operation agent and environment agent in the system, interaction rules between the agents and between the agent and environment were defined to describe the system’ s interrelation, a multi-agent coordination algorithm was given, and a joint firefighting and rescue simulation model was built. At last, a simulation is made on SWARM software to validate the proposed model.

财务风险的防范与控制是企业加强管理的重要内容。针对企业财务风险,本文将多Agent引入风险管理,通过对企业财务风险事件按性质进行分类,构建了财务风险管理Agent,并通过管理Agent和辅助Agent,组成多Agent系统。最后,本文构建了基于多Agent分布式架构的企业财务风险管理模型,并给出解决方案。

It is important to prevent and control financial risk for strengthening enterprises’ management . Using Multi-Agent system and classifying enterprise financial risk events according to the nature ,financial risk management agent was constructed and Multi-Agent system was formed together with the management agent and assisted agent.At last,enterprise financial risk management model was constructed based on Multi-Agent distributed architecture,and the solutions were given.

为了满足学校办学的智能化需求，设计了基于免疫多Agent协同的教育管理系统模型。研究了免疫多Agent协作处理模型，通过对话式扩充合同网协议完成了多Agent协作。采用免疫多Agent的协同进化算法提高了免疫进化算法的全局搜索能力和收敛速度，使教育管理系统模型具有良好的最优解搜索能力。

In order to meet the demand of intelligent design, school education management system model based on immune multi-agent collaboration was designed. Immune agent cooperation processing model was studied and multi-agent cooperation was constructed through the dialogue type extended contract net protocol. Coevolutionary algorithm of immune multi-agent enhanced the global search ability and convergence speed compared with the immune evolutionary algorithm, which makes the education management system model to have optimal search capability.