Hierarchical Planning Knowledge for Refining Partial-Order Plans

Hierarchical Planning Knowledge for Refining Partial-Order Plans. Lee-Urban, S. M. Ph.D. Thesis, Lehigh University, 2012.

Dissertation abstract bibtex

Automated plan synthesis, or “generative planning”,is a process whereby a course of action – a plan – is generated to achieve some desired goals. When the planning process uses a previously formulated plan as a starting point for solving a new problem, the process is called “plan adaptation”.Over the years there has been a significant research effort on plan adaptation. Part of the reason for this continuing interest in plan adaptation is attributable to studies indicating a wide range of potential applications, which include military planning, computer gaming, narrative computing, manufacturing, route planning, and medicine. Despite this remarkable body of research, existing plan adaptation algorithms do not scale well with problem size; even on medium-size problems, the performance of plan adaptation tends to be rather poor. Furthermore, for some worst-case situations it has been proven that the computational complexity of plan adaptation can be greater than that of generative planning. Although these worst-case scenarios have been shown to be inapplicable to most existing adaptation algorithms, it is nevertheless a fact that the lack of scalability of plan adaptation techniques is a major hurdle preventing their use in real-world applications.

Them main goal of this dissertation is to address the problem of how to efficiently represent and apply high-quality, hand-crafted plan refinement and plan adaptation knowledge. To accomplish this goal I studied domain-configurable plan adaptation, the results of which are presented in this dissertation. This new problem-solving paradigm uses domain-specific knowledge about plan adaptation to guide a domain-independent 2plan adaptation algorithm. This new technique is novel in that existing approaches for plan adaptation modify the input plan by either using domain-independent plan adaptation knowledge in a domain-independent adaptation algorithm or domain-specific plan adaptation knowledge in a domain-specific adaptation algorithm.By focusing on the use of hand-crafted expert partial-order planning control knowledge and its representation, one most notably eases the burden of knowledge acquisition –that is one does not need a complete knowledge base in order to create good plans. This permits the knowledge engineer to focus on encoding strategies most relevant to solving the problems, while leaving the more mundane and human-difficult chores of “plan bookkeeping” to the underlying planner.

@PhdThesis{Lee-Urban2012PhDThesis,
  author   = {Stephen Montgomery Lee-Urban},
  title    = {Hierarchical Planning Knowledge for Refining Partial-Order Plans},
  school   = {Lehigh University},
  year     = {2012},
  abstract = {Automated plan synthesis, or “generative planning”,is a process whereby a course of action -- a plan -- is generated to achieve some desired goals. When the planning process uses a previously formulated plan as a starting point for solving a new problem, the process is called “plan adaptation”.Over the years there has been a significant research effort on plan adaptation. Part of the reason for this continuing interest in plan adaptation is attributable to studies indicating a wide range of potential applications, which include military planning, computer gaming, narrative computing, manufacturing, route planning, and medicine. Despite this remarkable body of research, existing plan adaptation algorithms do not scale well with problem size; even on medium-size problems, the performance of plan adaptation tends to be rather poor. Furthermore, for some worst-case situations it has been proven that the computational complexity of plan adaptation can be greater than that of generative planning. Although these worst-case scenarios have been shown to be inapplicable to most existing adaptation algorithms, it is nevertheless a fact that the lack of scalability of plan adaptation techniques is a major hurdle preventing their use in real-world applications.<br/><br/>

  Them main goal of this dissertation is to address the problem of how to efficiently represent and apply high-quality, hand-crafted plan refinement and plan adaptation knowledge. To accomplish this goal I studied domain-configurable plan adaptation, the results of which are presented in this dissertation. This new problem-solving paradigm uses domain-specific knowledge about plan adaptation to guide a domain-independent 2plan adaptation algorithm. This new technique is novel in that existing approaches for plan adaptation modify the input plan by either using domain-independent plan adaptation knowledge in a domain-independent adaptation algorithm or domain-specific plan adaptation knowledge in a domain-specific adaptation algorithm.By focusing on the use of hand-crafted expert partial-order planning control knowledge and its representation, one most notably eases the burden of knowledge acquisition –that is one does not need a complete knowledge base in order to create good plans. This permits the knowledge engineer to focus on encoding strategies most relevant to solving the problems, while leaving the more mundane and human-difficult chores of “plan bookkeeping” to the underlying planner.},
  url_Dissertation = {https://preserve.lehigh.edu/cgi/viewcontent.cgi?article=2213&context=etd}
}

Downloads: 0

{"_id":"bYqFnNo9mFRF2gjjb","bibbaseid":"leeurban-hierarchicalplanningknowledgeforrefiningpartialorderplans-2012","authorIDs":[],"author_short":["Lee-Urban, S. M."],"bibdata":{"bibtype":"phdthesis","type":"phdthesis","author":[{"firstnames":["Stephen","Montgomery"],"propositions":[],"lastnames":["Lee-Urban"],"suffixes":[]}],"title":"Hierarchical Planning Knowledge for Refining Partial-Order Plans","school":"Lehigh University","year":"2012","abstract":"Automated plan synthesis, or “generative planning”,is a process whereby a course of action – a plan – is generated to achieve some desired goals. When the planning process uses a previously formulated plan as a starting point for solving a new problem, the process is called “plan adaptation”.Over the years there has been a significant research effort on plan adaptation. Part of the reason for this continuing interest in plan adaptation is attributable to studies indicating a wide range of potential applications, which include military planning, computer gaming, narrative computing, manufacturing, route planning, and medicine. Despite this remarkable body of research, existing plan adaptation algorithms do not scale well with problem size; even on medium-size problems, the performance of plan adaptation tends to be rather poor. Furthermore, for some worst-case situations it has been proven that the computational complexity of plan adaptation can be greater than that of generative planning. Although these worst-case scenarios have been shown to be inapplicable to most existing adaptation algorithms, it is nevertheless a fact that the lack of scalability of plan adaptation techniques is a major hurdle preventing their use in real-world applications.<br/><br/> Them main goal of this dissertation is to address the problem of how to efficiently represent and apply high-quality, hand-crafted plan refinement and plan adaptation knowledge. To accomplish this goal I studied domain-configurable plan adaptation, the results of which are presented in this dissertation. This new problem-solving paradigm uses domain-specific knowledge about plan adaptation to guide a domain-independent 2plan adaptation algorithm. This new technique is novel in that existing approaches for plan adaptation modify the input plan by either using domain-independent plan adaptation knowledge in a domain-independent adaptation algorithm or domain-specific plan adaptation knowledge in a domain-specific adaptation algorithm.By focusing on the use of hand-crafted expert partial-order planning control knowledge and its representation, one most notably eases the burden of knowledge acquisition –that is one does not need a complete knowledge base in order to create good plans. This permits the knowledge engineer to focus on encoding strategies most relevant to solving the problems, while leaving the more mundane and human-difficult chores of “plan bookkeeping” to the underlying planner.","url_dissertation":"https://preserve.lehigh.edu/cgi/viewcontent.cgi?article=2213&context=etd","bibtex":"@PhdThesis{Lee-Urban2012PhDThesis,\n author = {Stephen Montgomery Lee-Urban},\n title = {Hierarchical Planning Knowledge for Refining Partial-Order Plans},\n school = {Lehigh University},\n year = {2012},\n abstract = {Automated plan synthesis, or “generative planning”,is a process whereby a course of action -- a plan -- is generated to achieve some desired goals. When the planning process uses a previously formulated plan as a starting point for solving a new problem, the process is called “plan adaptation”.Over the years there has been a significant research effort on plan adaptation. Part of the reason for this continuing interest in plan adaptation is attributable to studies indicating a wide range of potential applications, which include military planning, computer gaming, narrative computing, manufacturing, route planning, and medicine. Despite this remarkable body of research, existing plan adaptation algorithms do not scale well with problem size; even on medium-size problems, the performance of plan adaptation tends to be rather poor. Furthermore, for some worst-case situations it has been proven that the computational complexity of plan adaptation can be greater than that of generative planning. Although these worst-case scenarios have been shown to be inapplicable to most existing adaptation algorithms, it is nevertheless a fact that the lack of scalability of plan adaptation techniques is a major hurdle preventing their use in real-world applications.<br/><br/>\n\n Them main goal of this dissertation is to address the problem of how to efficiently represent and apply high-quality, hand-crafted plan refinement and plan adaptation knowledge. To accomplish this goal I studied domain-configurable plan adaptation, the results of which are presented in this dissertation. This new problem-solving paradigm uses domain-specific knowledge about plan adaptation to guide a domain-independent 2plan adaptation algorithm. This new technique is novel in that existing approaches for plan adaptation modify the input plan by either using domain-independent plan adaptation knowledge in a domain-independent adaptation algorithm or domain-specific plan adaptation knowledge in a domain-specific adaptation algorithm.By focusing on the use of hand-crafted expert partial-order planning control knowledge and its representation, one most notably eases the burden of knowledge acquisition –that is one does not need a complete knowledge base in order to create good plans. This permits the knowledge engineer to focus on encoding strategies most relevant to solving the problems, while leaving the more mundane and human-difficult chores of “plan bookkeeping” to the underlying planner.},\n url_Dissertation = {https://preserve.lehigh.edu/cgi/viewcontent.cgi?article=2213&context=etd}\n}\n\n\n","author_short":["Lee-Urban, S. M."],"key":"Lee-Urban2012PhDThesis","id":"Lee-Urban2012PhDThesis","bibbaseid":"leeurban-hierarchicalplanningknowledgeforrefiningpartialorderplans-2012","role":"author","urls":{" dissertation":"https://preserve.lehigh.edu/cgi/viewcontent.cgi?article=2213&context=etd"},"metadata":{"authorlinks":{}}},"bibtype":"phdthesis","biburl":"hierarchical-task.net/bibtex/theses-in-hierarchical-planning.bib","creationDate":"2020-04-25T05:04:40.405Z","downloads":0,"keywords":[],"search_terms":["hierarchical","planning","knowledge","refining","partial","order","plans","lee-urban"],"title":"Hierarchical Planning Knowledge for Refining Partial-Order Plans","year":2012,"dataSources":["YjidmhgCHYRtoZTo6","vY5emPje84NjvCwk9"]}