Adaptive Constraint-Based Agents in Artificial Environments

[MAIN] [Goals]   [Software]   [Documentation]   [Publications]   [FAQ]   [Contact Information]

[ Please note: The project has been discontinued as of May 31, 2005 and is superseded by the projects of the ii Labs. There won't be further updates to these pages. ]

The following publications/drafts are available as PostScript/PDF files. The HTML sections are mostly more detailed, but do not cover all papers.


The Agent Model

  • Open World Planning as SCSP.
    In Papers from the AAAI-2000 Workshop on Constraints and AI Planning, Technical Report, WS-00-02, 35-46. AAAI Press, Menlo Park, California, 2000.

    The planning model for the EXCALIBUR agents is given here. For the handling of further aspects of incomplete knowledge, please have a look at the publication "A Planning Model for Agents in Dynamic and Uncertain Real-Time Environments" below.

  • A Planning Model for Agents in Dynamic and Uncertain Real-Time Environments.
    In Proceedings of the 1998 AIPS Workshop on Integrating Planning, Scheduling and Execution in Dynamic and Uncertain Environments, Technical Report, WS-98-02, 7-14. AAAI Press, Menlo Park, California, 1998.

    This document describes the basic planning model for EXCALIBUR's agents. The formal model (slightly revised) is described in the publication "Open World Planning as SCSP" (see above).

  • Planning to Plan - Integrating Control Flow.
    In Proceedings of the International Workshop on Heuristics (IWH'02), 79-84, 2002. Also appeared in Tsinghua Science and Technology 8(1): 1-7, 2003.

    The publication describes how an agent can reason about its own reasoning process. This is necessary for tasks like negotiation or for explicitly reasoning about ways to optimize its planning process.

  • Planning in Dynamic Worlds: More Than External Events.
    In Proceedings of the IJCAI-03 Workshop on Agents and Automated Reasoning, 30-35.

    An agent must not only consider the impact of his own actions but also how the world around him evolves. The publications describes how we approach this problem for our agent architecture.

An Agent's Reasoning

  • Using Global Constraints for Local Search.
    In Freuder, E. C., and Wallace, R. J. (eds.), Constraint Programming and Large Scale Discrete Optimization, American Mathematical Society Publications, DIMACS Volume 57, 9-28, 2001.

    This document describes our base technology to allow for anytime properties and the integration of domain-dependent knowledge.

  • Local-Search Heuristics for Generative Planning.
    In Proceedings of the Fifteenth Workshop on AI in Planning, Scheduling, Configuration and Design (PuK 2001), 56-70, 2001.

    This paper presents the local-search heuristics applied in the EXCALIBUR agent's planning system and demonstrates an application to example problems.

  • Choosing Search Heuristics by Non-Stationary Reinforcement Learning.
    In Resende, M. G. C., and de Sousa, J. P. (eds.), Metaheuristics: Computer Decision-Making, Kluwer Academic Publishers, 523-544, 2003.

    The article compares different selection schemes to choose from a number of alternatives. These strategies are applied in the EXCALIBUR agent's planning system for plan generation.

  • A Modular Graphical User Interface for Interactive Planning.
    Diploma Thesis, Technical University of Munich, Department of Computer Science, 2004.

    The thesis describes the development and functionality of the graphical planning interface that we use to visualize (and debug) the planning process.

Formal Concepts

  • Constraints and AI Planning.
    IEEE Intelligent Systems, 20(2), 62-72, 2005.

    A survey on constraint-based approaches that can be used for an agent's action planning. Our own approach, based on so-called structural constraint satisfaction (see below for more details), is briefly described as well.

  • Structural Constraint Satisfaction.
    In Papers from the 1999 AAAI Workshop on Configuration, Technical Report, WS-99-05, 76-82. AAAI Press, Menlo Park, California, 1999.

    This document introduces the concept of structural constraint satisfaction, which is necessary to handle a variable plan structure. However, it is based on refinement search. The paper "Applying Local Search to Structural Constraint Satisfaction" (see below) extends this concept to EXCALIBUR's local search approach.

  • Applying Local Search to Structural Constraint Satisfaction.
    In Proceedings of the 1999 IJCAI Workshop on Intelligent Workflow and Process Management: The New Frontier for AI in Business, 1999.

    The concept of structural constraint satisfaction is combined with local search and global constraints. This forms EXCALIBUR's base technology for an agent's behavior planning.

  • Realisierung struktureller Constraints.
    Diploma Thesis, Technical University of Berlin, Department of Electrical Engineering and Computer Science, 2003.

    The thesis explains how the handling of structural constraints was realized for our engine. Because of the complexity of this reasoning, however, we will most likely use this feature only for internal testing/prototyping. For the externally available planning process, we will ensure that heuristics do not make changes that lead to inconsistent graphs. Anyone who is interested in using the feature of structural constraints (e.g., when extending or changing our planning model) can activate it in a very simple way.

[MAIN] [Goals]   [Software]   [Documentation]   [Publications]   [FAQ]   [Contact Information]

For questions, comments or suggestions, please contact us.

Last update:
March 23, 2005 by Alexander Nareyek