机构地区: 暨南大学信息科学技术学院计算机科学系
出 处: 《计算机应用与软件》 2005年第5期16-18,共3页
摘 要: 图规划算法在智能规划的研究和发展中具有重要的意义和价值,它采用图的方式来求解规划问题,并且提出了用于规划的“规划图”的概念[1,2]。这一方法可以获得经典规划问题的最少动作数意义的最优解,但是却不能求解与动作代价相关的数值规划问题。本文提出一种带权值的图规划算法,在规划图的“动作增加效果”边上增加权值,用来表示动作产生这些效果的代价,从而使得通过规划图来获得最小动作代价意义的最优解。因此,这种方法不但提高了规划解的质量,而且所得到的规划更加贴近实用。 The GraphPlan algorithm has a very significant effect on the research and de velopment of AI planning,which first proposed the concept of planning-graph and solves classic planning problems based on this special data-structure.Surely,t he method can obtain a planning solution with least action amount,however,it has a difficulty on figuring out the numeric planning problems that often involves the action cost.So we propose a weighted algorithm of GraphPlan,which adds the w eight,that is,the cost that the action makes its effects true,into the “action -add-effect”border in the planning-graph,then we can get a planning solution with minimal action cost.Therefore,this method not only improves the quantity o f planning solution,but also makes it more practical.
关 键 词: 规划算法 规划问题 规划图 智能规划 最优解 代价 加权值 求解 数值 最小
领 域: [自动化与计算机技术] [自动化与计算机技术] [理学] [理学]