实时专家系统中反应—规划—动作的集成  

Integration of Reacting-Planning-Execution in Realtime Expert Systems

作　　者：郭宏飞[1] 李景银[1] 许桂清[1] 周建常[1] 

机构地区：[1]东北大学信息科学与工程学院,沈阳110005

出　　处：《控制与决策》1997年第A00期462-466,476,共6页Control and Decision

摘　　要：以AMPE模型为基础,探讨了在动态环境下AI系统如何集成响应执行和规划生成的问题。AMPE模型是从实际应用中演化出的AI模型,采用分离结构,在规划中引入一类被称为观察点的特殊行为,支持异步重规划。描述了AMPE的各类规划行为的形式和消息的基本结构。给出一个规划表示的实例。Reacting, planning and execution are essential capabilities of AI systems situated in dynamic envi-ronments. The paper presents an asynchronous model of integrating planning and execution (AMPE) in a complex dynamic environment. AMPE is evolved from some real -time AI applications. In AMPE , Obser-vation Point which represents a kind of special plan action is introduced into plan representation. The Obser-vation Point provides a window from which the planner can observe the execution of the plan. The asyn-chronous replanning capability is supported too. The paper addresses how to reconcile reacting, planning and execution in a complex dynamic environment. Representations of basic planning actions and structures of ba-sic messages are described. An example of plan is described.

关 键 词：人工智能 实时专家系统 规划表示 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]