公开(公告)号：KR1020070061260A

公开(公告)日：2007-06-13

申请号：KR1020060076669

申请日：2006-08-14

Applicant: 한국전자통신연구원

Inventor： 김민경 ,  김현

IPC: G06F15/18 ,  G06F9/48

CPC classification number: G06F9/4831

Abstract: An intelligent multi-agent system through a learning engine and an operation method thereof are provided to apply the learning engine to maximize context recognition dynamically changed in a ubiquitous environment and apply a learning result to interaction among various kinds of agents. Each zone agent(4) is in each area, manages user state information, and performs a service corresponding to a generated event. The learning engine(1) observes/learns a user behavior pattern of each zone agent and outputs the learned behavior pattern in a rule form. A task generator(3) generates a task to the zone agent when the new rule is generated. An environment agent(5) provides zone agent environment information to the task generator to refer to zone information of the zone agent. A rule database(2) stores the rules output from the learning engine between the learning engine and the task generator, and stores a permanent definition rule.

Abstract translation: 提供了一种通过学习引擎及其操作方法的智能多代理系统，以应用学习引擎来最大化在无所不在的环境中动态改变的语境识别，并将学习结果应用于各种代理之间的交互。 每个区域代理（4）在每个区域中，管理用户状态信息，并执行与生成的事件相对应的服务。 学习引擎（1）观察/学习每个区域代理的用户行为模式，并以规则形式输出学习的行为模式。 任务生成器（3）在生成新规则时向区域代理生成任务。 环境代理（5）向任务生成器提供区域代理环境信息以引用区域代理的区域信息。 规则数据库（2）将从学习引擎输出的规则存储在学习引擎和任务生成器之间，并存储永久定义规则。