Classical planners are deficient in dealing with incomplete knowledge, resulting in incorrect interpretation of situations. This paper introduces a new planner ACP that tackles the problem by combining an action model based on Brown's quantified modal logic Z and a bi-directional reasoning scheme. The commonsense reasoning is handled by doing case analysis on default formulas that characterize each action's logical meaning. ACP suggests a simple and efficient approach for maintaining and updating consistent beliefs at major action steps. The frame problem and the qualification problem are dealt by introducing appropriate default rules and consistency criterion to the action model. Incremental constraints are used to ensure that the next state will be consistent with achieving the goals. The whole planning process can be modified when an unexpected action outcome occurs. The research also shows that if a plan is logically possible, ACP will find it. Therefore, ACP is complete.
古典计划者缺乏处理不完整知识的能力,导致对情况的错误解释。本文介绍了一种新的计划程序ACP,它通过结合基于布朗的量化模态逻辑Z的动作模型和双向推理方案来解决该问题。常识推理是通过对表征每个动作的逻辑含义的默认公式进行案例分析来处理的。 ACP提出了一种简单有效的方法来维持和更新主要行动步骤上的一致信念。通过向动作模型引入适当的默认规则和一致性准则来解决框架问题和资格问题。增量约束用于确保下一个状态与实现目标一致。当发生意外的行动结果时,可以修改整个计划过程。研究还表明,如果计划在逻辑上可行,ACP会找到它。因此,ACP已完成。 P>
Department of Computer Science, University of Kansas, Lawrence, KS;
机译:影响患者患者患者痔疮的因素:基于计划行为理论和常识模型的研究
机译:应对英格兰规划和发展的保守常识
机译:具有常识性推理的分层任务网络规划,用于多人行为分析
机译:从文本中提取常识知识以进行机器人计划
机译:不只是常识:在两所中学中,原则性地理解和实施“共同核心”。
机译:了解乳腺癌的手术选择:计划行为理论和常识模型如何有助于决策支持干预措施?
机译:区域规划:共同感觉还是非常规?
机译:战术林业规划。计算机辅助的常识,以及与战略计划相关的模型