Extending Graphplan to Handle Uncertainty and Sensing Actions

扩展 Graphplan 来处理不确定性和感知动作

• 批准号: 9872128
• 负责人: Daniel Weld
• 金额: $ 23.7万
• 依托单位: University of Washington
• 依托单位国家: 美国
• 项目类别: Standard Grant
• 财政年份: 1998
• 资助国家: 美国
• 起止时间: 1998-09-15 至 2002-12-31
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=9872128&HistoricalAwards=false
• 关键词: Extending; Graphplan; Handle; Uncertainty; Sensing

This research is in extending the Graphplan algorithm, which uses disjunctive representation and local constraint satisfaction techniques to handle a wider variety of planning, rather than being restricted to theclassical planning assumptions of determinism and complete information. Extensions are to allow contingent planning, probabilistic planning, expressive sensing actions, interleaved planning and execution, andreal-valued durations. The same methodology that characterizes the PI's previous research are being used: a balance between theoretical rigor (proofs of soundness and completeness) and empirical evaluation ofimplementation performance. Source code of the implementations will be distributed, and it is anticipated that better, more versatile fast planning algorithms will result.http://www.cs.washington.edu/homes/weld/weld.html

这项研究是对 Graphplan 算法的扩展，该算法使用析取表示和局部约束满足技术来处理更广泛的规划，而不是局限于确定性和完整信息的经典规划假设。 扩展允许应急规划、概率规划、表达感知行动、交错规划和执行以及实值持续时间。 PI 之前的研究采用了相同的方法：理论严谨性（合理性和完整性的证明）与实施绩效的实证评估之间的平衡。 实现的源代码将被分发，预计将产生更好、更通用的快速规划算法。http://www.cs.washington.edu/homes/weld/weld.html