摘要
基于模型检测的规划方法是最近发展起来的新方法,它可以处理带有不确定性的规划问题.分别设计了对求弱规划解、强规划解和强循环规划解的问题中的状态进行分层的方法.状态被分层后,求规划解只需要在从上层到其下一层状态之间寻找状态动作序偶就可以了,其他状态动作序偶都可以去掉.分别获得了求弱规划解、强规划解和强循环规划解时状态被分层后的一些重要性质,这些性质是关于一些状态动作序偶是否可以不参与构成弱规划解、强规划解和强循环规划解的结论.通过所获得的性质可以将大量的状态动作序偶直接去掉,从而减少问题规模.以往的对基于模型检测规划的研究都是采用从目标状态开始的反向搜索方法,在状态被分层以后可以采用正向搜索技术展开相应的研究.
Planning by model checking is an approach to planning under uncertainty that deals with nondeterminism. Three ways which obtain hierarchical states for searching weak planning, strong planning, and strong cyclic planning are respectively designed. Based on hierarchical states, some important conclusions on a weak solution, a strong solution, and a strong cyclic solution are obtained. What can be eliminated directly are all found when a weak solution, a strong solution, and a strong cyclic solution are in turn searched. Therefore many state-action pairs can be eliminated directly before starting planning. In fact, a way has been given which is based on a search proceeding forwards from the initial states towards the goal states.
出处
《软件学报》
EI
CSCD
北大核心
2009年第4期858-869,共12页
Journal of Software
基金
国家自然科学基金Nos.60673193,60773047,60773201
湖南省重点学科建设项目No.081202
湖南省教育厅科研项目No.08C874
湘潭大学校基金No.kz08009~~
关键词
模型检测
状态分层
不确定规划
正向搜索
状态动作序偶
model checking
hierarchical state
planning under uncertainty
forward search
state-action pair