Logic-based Knowledge Representation

Title

Principal investigators

Funding agency

Deutsche Forschungsgemeinschaft (DFG), Project TH 541/14

Project abstract

The aim of this project is to establish a direct cooperation between researchers in Action Formalisms, Description Logic, Default Reasoning and Planning so that they may reap the benefits of the combined advances both in theory and practice in each of the respective areas. The overall goal is to achieve a tight integration of state-of-the-art systems from each area. Thus, the project consists of the following steps:

Relating Situation Calculus and Fluent Calculus

The two action formalisms fluent calculus and situation calculus form the crucial link of the project. Thus establishing a tight relation between the two is fundamental. Ideally, this involves obtaining faithful translations of domain axiomatizations in one of the two calculi into the other. Furthermore, an effective method for translating between programs for Flux and Golog, the respective leading implementations, is to be develloped.

Action Formalisms with Description Logic

In this project we study fragments of the fluent calculus obtained by restricting parts of its underlying logic from full first order predicate logic to suitable description logics so that certain reasoning tasks will be decidable. A complementary goal is the investigation of how description logic can benefit from integrating ideas of the reasoning about action community, e.g. for reasoning about semantic web services or modelling dynamic domains in the life sciences. Devising and implementing a logic programming language on top of these these insights is a major objective of this project.

Planning Techniques and Action Languages

The goal of this project is the development of a common semantics for (a fragment of) situation calculus and the planning domain description language PDDL. This is to enable the integration of state-of-the-art planning systems into action formalisms. In the other direction planning systems could benefit from modern concepts in action languages like e.g. continuous and concurrent actions.

Defaults and Preferences in Action Formalisms

Researchers in Nonmonotonic Logics have developed a rich repertoire of methods for drawing default conclusions - that is, conclusions that are plausible in the light of the available information. Integrating these methods into a general action formalism is the goal of this project. This will increase the expressivity of action calculi in that likely - as opposed to certain - outcomes of actions or preferred strategies can naturally be modeled.