PARIS: Plan Abstraction and Refinement in an Integrated System

Within PARIS, case-based action planning using abstraction and explanation-based techniques is investigated in detail. Traditionally, case-based reasoning approaches retrieve, reuse, and retain cases given in a single, concrete representation. PARIS is a domain independent case-based planning system that differs from this traditional approach in that it introduces abstraction techniques into the case-based reasoning process. PARIS retrieves, reuses and retains cases at different (higher) levels of abstraction. In a nutshell, PARIS works as follows. Available planning cases given at the concrete level are abstracted to several levels of abstraction which leads to a set of abstract cases that are stored in the case-base. Case abstraction is done automatically in the retain phase of the CBR-cycle. When a new problem must be solved, an abstract case is retrieved whose abstract problem description matches the current problem at an abstract level. In the subsequent reuse phase, the abstract solution is refined, i.e., the details that are not contained in the abstract case are added to achieve a complete solution of the problem. This refinement is done by a generative planner that performs a forward directed state space search. Besides case abstraction and refinement, PARIS also includes an explanation-based approach for generalizing cases during learning and for specializing them during problem solving. This technique allows to further increase the flexibility of reuse.

The Paris Architecture


 

Funding

Funding by the University of Kaiserslautern from 1990 - 1996.

Research Team

 

Publications

1998

  • Ralph Bergmann, Hector Munoz-Avila, M. Veloso, and E. Melis. Case-based reasoning applied to planning tasks. In M. Lenz, B. Bartsch-Spörl, H.-D. Burkhard, and Stefan Wess, editors, Case-Based Reasoning Technology from Foundations to Applications, 1998, Springer-Verlag.
  • Ralph Bergmann. Efficient retrieval of abstract cases for case-based planning. In Ellman, editor, Proceedings on Symposium on Abstraction, Reformulation and Approximation (SARA'98), pages 9-18, Pacific Grove, California, 1998.

1996

  • Ralph Bergmann. Effizientes Problemlösen durch flexible Wiederverwendung von Fällen auf verschiedenen Abstraktionsebenen , number 138 in DISKI, Infix, 1996.
  • Ralph Bergmann, and Wolfgang Wilke. PARIS: Flexible plan adaptation by abstraction and refinement. In ECAI 1996 Workshop on Adaptation in Case-Based Reasoning, 1996.

1995

  • Ralph Bergmann, and Wolfgang Wilke. Building and refining abstract planning cases by change of representation language. Journal of AI Research, , 1995.
  • Ralph Bergmann, and Wolfgang Wilke. Learning abstract planning cases. In Proceedings of the Europ. Conf. on Machine Learning, pages 55-76, 1995, Springer Verlag.
  • Ralph Bergmann, and Wolfgang Wilke. Flexible Reuse of Plans by Abstraction and Refinement. In Proceedings of the IJCAI Workshop on Reuse of Plans, Proofs, and Programs, 1995.
  • Ralph Bergmann. Plan Abstraction with Change of Representation Language. In Proceedings of the Symposium on Abstraction Reformulation and Approximation (SARA'95), 1995.

 1994

  •   Ralph Bergmann, Gerhard Pews, and Wolfgang Wilke. Explanation-based similarity: A unifying approach for integrating domain knowledge into case-based reasoning for diagnosis and planning tasks,Topics in Case-Based Reasoning Lecture Notes in AI. Springer-Verlag, 1994.
  • Ralph Bergmann, and Wolfgang Wilke. Lernen von Abstraktionshierarchien zur Optimierung der Auswahl von maschinell abstrahierten Plaenen. In Proceedings des 8th Workshop "Planen und Konfigurieren" (PuK-94) der GI, Universitaet Kaiserslautern, 1994.

1993

  •   Ralph Bergmann. Integrating Abstraction, Explanation-based Learning from Multiple Examples, and Hierarchical Clustering with a Performance Component for PlanningIn. In Enric Plaza, editor, Proc. ECML-93 Workshop on "Integrated Learning Architectures", 1993.
  • Ralph Bergmann, and Wolfgang Wilke. Inkrementelles Lernen von Abstraktionshierarchien aus maschinell abstrahierten Plänen. In Proceedings des Workshops: Maschinelles Lernen: Theoretische Ansätze and Anwendungsaspekte, Forschungsberichte Universität Karlsruhe, Bericht 291, 1993.
  • Gerhard Pews, and Ralph Bergmann. Erklaerungsbasierte Aehnlichkeitsbestimmung und ihre Anwendung in Planung und Diagnose. In Proceedings des Workshops: Maschinelles Lernen: Theoretische Ansätze and Anwendungsaspekte, Forschungsberichte Universität Karlsruhe, Bericht 291, 1993.
  • Ralph Bergmann. Learning Hierarchically Clustered Shared Plan Abstractions as Problem solving Knowledge with High utility for planning. In Proceedings des 7th Workshop "Planen und Konfigurieren" (PuK-93) der GI. Arbeitspapiere der GMD, 1993.

1992

  • Ralph Bergmann. Knowledge acquisition by generating skeletal plans, Contemporary Knowledge Engineering and Cognition. Springer-Verlag, 1992.
  • Ralph Bergmann, and A. Dannenmann. Abstrakte Fallrepraesentation fuer Abruf und Adaption von Planungsfaellen auf verschiedenen Abstraktionsniveaus. In Workshop: Aehnlichkeit von Faellen in Systemen des fallbasierten Schliessens, 1992.
  • Ralph Bergmann. Learning Plan abstractions. In GWAI-92 16th German Workshop on Artificial Intelligence, volume 671 of LNAI, 1992, Springer.
  • Ralph Bergmann. Learning abstract planning cases to speed up hierarchical planning. In Proceedings of the ML'92 Workshop on Knowledge Compilation and Speedup Learning, 1992.