抽象层次上FO—POMDP的引入,使得人们可简洁地、陈述地表达复杂的POMDP,解决常规POMDP在实际中所无法解决的大规模决策问题。介绍了FO—POMDP的基础,包括状况表达式、行动、观察值和观察函数。提出了一阶信念状态的概念,并分别针对随机转移行动和随机观察行动给出一阶信念状态的更新方法。最后用FO—Tiger-Grid模型对一阶信念状态的概念和更新方法进行了实例分析验证。%Using FO-POMDP on abstract level, one can compactly and declaratively represent complex POMDP. And one can solve real world problems that generally have large state space with regular POMDP. This paper introduces the concept of FO- POMDP, including the expression of case, stochastic actions, and the concept of observation function. Then the concept of first-order belief state is given. The method of updating the first-order belief state is provided, which is the important contribution of this paper. Finally, method is tested by FO-Tiger-Grid.
展开▼