一种决策论和人工智能相结合的非确定性规划方法

A Method of Planning under Uncertainty by Combining AI and Decision Theory

一、前言规划是人工智能中的一个重要分支。一个规划是一个动作序列,其执行的结果将世界从初始状态带到目标状态。在规划研究的历史中,一个重大的突破是七十年代初的STRIPS,它在解决“画面”问题(frame problem)中进行了成功的尝试。在STRIPS中,一个动作由一组条件、一个增加表和一个删除表描述。动作的后果包含在这两个表中。任何命题或谓词,只要不在这两个表中。

This paper proposes an approach to planning under uncertainty by combining artificial intelligence and decision theoretic methods. Goals are defined in decision theoretic framework. Actions are selected by Maximum Subjective Expected Utility rule. A causal model is established to compute probability distribution over world states. A planning procedure is given.