About | Study | Members | Projects | Publications | Events | Search | Internals | Contact & Imprint
Logic-based Knowledge Representation - Action Formalisms with Description Logic
Title
Action Formalisms with Description Logic
Principal investigators
Funding agency
Deutsche Forschungsgemeinschaft (DFG), Project BA 1122/13
Project abstract
The aim of this project is the integration of description logic and action formalisms. The motivation for this integration is twofold.
On the one hand, general action calculi like the fluent calculus and the situation calculus are based on full first order logic. This entails undecidability in general of basic reasoning tasks like e.g. checking state consistency, action applicability or computing updated states. By identifying suitable description logics for describing the current world state these issues may be adressed.
On the other hand, the need for integrating some kind of action representation into description logics has arisen. Description logics are a highly successful static knowledge representation formalism with applications e.g. on the semantic web or in the life sciences. Clearly, it is desirable to have the means to model semantic web services or reason about dynamic domains in the life sciences, like e.g. clinical protocols.
In a first step structured and decidable world descriptions based on description logic shall be integrated into the fluent calculus. This includes the identification of a description logic that is suited well for describing current world states via ABoxes and the development of a method for updating these ABoxes. It will then have to be shown that the method is correctly captured by the fluent calculus axiomatisation of update.
In a second step the work shall be extended to updating ABoxes in the presence of general TBoxes. On the fluent calculus side this corresponds to updates with indirect effects, giving rise to the well-studied ramification problem. It will have to be examined whether existing solutions can be adapted to this particular setting. Another goal is to investigate whether less expressive but tractable description logics might also be suited for describing world states.
Another objective of this project is to develop a version of the logic programming paradigm designed specifically for programming intelligent agents. This may be thought of as adapting the successful constraint-logic programming scheme CLP(X) to CLP(Sigma), where Sigma is a domain axiomatization in an action calculus. Of course, for this it is of tantamount importance that the special atoms of the logic program can effectively be decided via the underlying domain axiomatization. The resulting scheme instantiated with the action calculi developed in the afore-mentioned steps can then be implemented by combining the mature technologies of both plain prolog and description logic reasoners. The system will be evaluated by modeling semantic web services or clinical protocols.
Start date
01.09.2005
Duration
48 months