MIME-Version: 1.0 Server: CERN/3.0 Date: Monday, 06-Jan-97 21:19:13 GMT Content-Type: text/html Content-Length: 8170 Last-Modified: Monday, 26-Aug-96 18:13:31 GMT Tara Estlin's Papers

Tara Estlin's Papers

To view a paper, click on the open book image.

  1. Multi-Strategy Learning of Search Control for Partial-Order Planning
    Tara A. Estlin and Raymond J. Mooney
    Proceedings of the Thirteenth National Conference on Aritificial Intelligence, pp. 843-848, Portland, OR, August, 1996. (AAAI-96)

    Most research in planning and learning has involved linear, state-based planners. This paper presents SCOPE, a system for learning search-control rules that improve the performance of a partial-order planner. SCOPE integrates explanation-based and inductive learning techniques to acquire control rules for a partial-order planner. Learned rules are in the form of selection heuristics that help the planner choose between competing plan refinements. Specifically, SCOPE learns domain-specific control rules for a version of the UCPOP planning algorithm. The resulting system is shown to produce significant speedup in two different planning domains.

  2. Integrating EBL and ILP to Acquire Control Rules for Planning
    Tara A. Estlin and Raymond J. Mooney
    Proceedings of the Third International Workshop on Multi-Strategy Learning, pp. 271-279, Harpers Ferry, WV, May 1996. (MSL-96).

    Most approaches to learning control information in planning systems use explanation-based learning to generate control rules. Unfortunately, EBL alone often produces overly complex rules that actually decrease planning efficiency. This paper presents a novel learning approach for control knowledge acquisition that integrates explanation-based learning with techniques from inductive logic programming. EBL is used to constrain an inductive search for selection heuristics that help a planner choose between competing plan refinements. SCOPE is one of the few systems to address learning control information in the newer partial-order planners. Specifically, SCOPE learns domain-specific control rules for a version of the UCPOP planning algorithm. The resulting system is shown to produce significant speedup in two different planning domains.

  3. Integrating Explanation-Based and Inductive Learning Techniques to Acquire Search-Control for Planning
    Tara A. Estlin
    Ph.D. proposal, Department of Computer Sciences, University of Texas at Austin, 1996.

    Planning systems have become an important tool for automating a wide variety of tasks. Control knowledge guides a planner to find solutions quickly and is crucial for efficient planning in most domains. Machine learning techniques enable a planning system to automatically acquire domain-specific search-control knowledge for different applications. Past approaches to learning control information have usually employed explanation-based learning (EBL) to generate control rules. Unfortunately, EBL alone often produces overly complex rules that actually decrease rather than improve overall planning efficiency. This paper presents a novel learning approach for control knowledge acquisition that integrates explanation-based learning with techniques from inductive logic programming. In our learning system SCOPE, EBL is used to constrain an inductive search for control heuristics that help a planner choose between competing plan refinements. SCOPE is one of the few systems to address learning control information for newer, partial-order planners. Specifically, this proposal describes how SCOPE learns domain-specific control rules for the UCPOP planning algorithm. The resulting system is shown to produce significant speedup in two different planning domains, and to be more effective than a pure EBL approach.

    Future research will be performed in three main areas. First, SCOPE's learning algorithm will be extended to include additional techniques such as constructive induction and rule utility analysis. Second, SCOPE will be more thoroughly tested; several real-world planning domains have been identified as possible testbeds, and more in-depth comparisons will be drawn between SCOPE and other competing approaches. Third, SCOPE will be implemented in a different planning system in order to test its portability to other planning algorithms. This work should demonstrate that machine-learning techniques can be a powerful tool in the quest for tractable real-world planning.

  4. Hybrid Learning of Search Control for Partial-Order Planning
    Tara A. Estlin and Raymond J. Mooney
    New Directions in AI Planning, M. Ghallab and A. Milani, Eds, IOS Press, 1996, pp. 129-140.

    This paper presents results on applying a version of the DOLPHIN search-control learning system to speed up a partial-order planner. DOLPHIN integrates explanation-based and inductive learning techniques to acquire effective clause-selection rules for Prolog programs. A version of the UCPOP partial-order planning algorithm has been implemented as a Prolog program and DOLPHIN used to automatically learn domain-specific search control rules that help eliminate backtracking. The resulting system is shown to produce significant speedup in several planning domains.

  5. Why Real-world Planning is Difficult: A Tale of Two Applications
    Steve Chien, Randall Hill, Jr., XueMei Wang, Tara Estlin, Kristina Fayyad, and Helen Mortenson
    New Directions in AI Planning, M. Ghallab and A. Milani, Eds, IOS Press, 1996, pp. 287-298.

    In this paper we describe a number of obstacles hampering the application of planning technology to real-world problems, as encountered in two real-world planning projects at JPL: MVP - a planning system for automated generation of image processing procedures; and LMCOA - an intelligent system for assistance in antenna operations. First, we describe how existing planning representation must be enhanced to represent and reason about aspects of plans besides goal achievement - resource usage, quality, execution time, flexibility, and generality. Second, planning systems must be able to fit into a wide range of operational contexts - most planning tasks cannot be completely automated, therefore at a minimum the plans produced must be easily understandable and modifiable by the users. In some cases the user must be intimately involved in the plan construction process itself. Third, planning systems must be able to compare favorably in terms of software lifecycle costs to other means of automation such as scripts or rule-based expert systems. This means that development of intelligent tools and environments to facilitate knowledge acquisition, validation, and maintenance are of prime importance. We hope that our description and elucidation of these issues will lead to increased work in these areas.