中文摘要：

通用规划（解）是针对某个领域的像算法一样的规划解,通过对其的解释可以直接得出具体问题的规划解,而不需要调用任何规划系统.但是目前通用规划的提取只能在一些简单或者特殊的领域中进行,没有推广到复杂或者一般的规划领域.该文提出在包含派生谓词的规划领域自动获取通用规划的方法.与已有获取方法不同的是：首先,基于派生谓词规则,文中方法明确指出派生谓词目标与动作效果之间的依赖关系,用以完善通用规划中动作应用的目的;其次,在提取过程中借助角色来帮助识别规划解中的循环结构.实验结果表明,文中方法不仅容易在派生谓词规划领域中获取通用规划,而且还能够以较好的性能求解一类以派生谓词为主要目标的规划“难”题.该文是在派生谓词规划领域中提取通用规划的首创性工作.

英文摘要：

A generalized plan is algorithm like and domain specific.One may achieve a solution plan directly by executing it,given a special planning problem,so that he does not need to rely on any general purpose underlying planner.However,at present generalized plans can be extracted only in some very simple or particular domains,and no work is involved in other complex or general ones.To break such limitations,this paper presents the method of acquiring automatically generalized plans for planning domains which include derived predicates.Unlike existing methods,our method explicitly indicates the dependency relationship between derived predicate goals and action effects,based on derived predicate rules.This would complement effectively intentions of action applications in a generalized plan.Also,our method takes full advantages of roles to help identify loop structures in plan examples.Experiment results on some test domains show that our method not only can be easily used to extract generalized plans in planning domains with derived predicates,but also has good performances in solving a kind of ＂hard＂ planning problem that have derived predicates as main goals.Our work takes the first step in acquiring automatically generalized plans in derived predicate planning domains.