Adaptive Constraint-Based Agents in Artificial Environments

[APPLICATION]   [Orc Quest Revisited]   [Domain-Independent Planning]   [Conclusion]

[ Please note: The project has been discontinued as of May 31, 2005 and is superseded by the projects of the ii Labs.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. ]

Domain-Independent Planning

(Related publication: [PUBLink])

The constraints used in the previous section work well for the Orc Quest example, but are not generally applicable. A preference value for each possible modification is not feasible for more complicated resources. Also, precondition requirements and temporal aspects were neglected. This section presents more general constraints, which can be used for a wider range of problems. Of course, the conceptual approach still makes it possible to easily integrate domain-specific knowledge for specific domains.

Sections [Global Action Resource Constraints] and [Global Task Constraints] already presented solutions for an Action Resource Constraint and a Task Constraint. They were developed for the job-shop scheduling domain but can be applied here as well. A small extension is made to the constraints, which is described in Section [Extending the ARC and TC].

The most important constraint missing for the planning domain is the State Resource Constraint. In Section [A State Resource Constraint with a Symbolic State Domain], a State Resource Constraint is presented that features a symbolic state domain (like the Location SRC of this figure). This enables most of the common benchmark domains to be handled, e.g., problems specified in the STRIPS formalism.

The resulting system is evaluated in Section [Evaluation].


[APPLICATION]   [Orc Quest Revisited]   [Domain-Independent Planning]   [Conclusion]

For questions, comments or suggestions, please contact us.

Last update:
May 20, 2001 by Alexander Nareyek