基于感知行动的一种双向规划算法研究

A Bi-direction Conditional Planning Algorithm Based on Sensing Actions

规划最主要的内容是对Agent行动进行推理,但Agent对真实世界环境的知识是不完备的,因此这种情况下的规划,需要感知行动以扩充Agent的知识状态,带有感知行动的规划,其结果不再是一个简单的行动序列,而是带有感知行动及其条件说明的条件规划。本文根据Agent知识状态空间的搜索方向,讨论了正向规划、反向规划,最后提出了一个能够处理感知行动和条件规划的双向规划算法BCPS(Bi-direction Conditional Planning with Sensing actions)。

The most important aspect of planning is reasoning about actions. But in reality agent＇s knowledge about environment is often incomplete. Then sensing actions are needed in these situations to collect information in cases it is needed to expand the agent＇s knowledge about environment. Then planning with incomplete information is no longer a simple sequence of actions, but a conditional plan with sensing actions. Based on the searching directions, the algorithms for planning obtained by means of progression searching, regression searching are discussed in details and a bi-direction planning algorithm（BCPS） is proposed in this paper.