<b>RAP</b> - <b>R</b>easoning <b>A</b>bout <b>P</b>lans Project RAP

DFKI

Von dieser Seite gibt es demnächst auch eine deutsche Fassung

RAP is a flexible architecture for the implementation of complex plan reasoning systems to be embedded into larger applications, like intelligent help systems, tutoring environments, simulation systems or general planning assistants. Besides plan generation and execution a variety of additional plan reasoning methods are needed in these contexts.

Implementing plan reasoning systems for this type of applications requires not only a wide spectrum of basic methods but also some flexible means to combine these methods and to modify existing systems conveniently.

RAP presents itself as a domain independent and flexible plan reasoning architecture consisting of three components:

The toolbox comprises reasoning services, like plan generation, plan recognition, plan library update and retrieval, plan modification, plan validation, temporal projection, plan verification, and probabilistic plan reasoning. These services are implemented as independent modules providing carefully specified interfaces (i/o channels) which basically enable a most flexible combination of these modules.

The shell has been implemented as a special kind of visual programming environment. Using this tool, plan reasoners are configured by graphically designing a so-called module net. The nodes of such a net are modules while the arcs, combining appropriate i/o channels, reflect the data flow between them. The graphical programming language provides features like conditional branching, iteration, and concurrency, thereby easing the design process of complex systems.

The domain modeling tool enables a domain specialist to set up a domain model without being too much troubled by the details of the system's internal representation formalism, and supportes him in keeping the domain model consistent.

The tool provides a graphical user interface with fixed definition schemata for domain objects, their properties, relations between them, domain constraints, actions, plans, and planning problems. Based on the definition schemata the system uniformly generates sets of modal logic axiom schemata and automatically proves that none of the actions violates any of the domain constraints specified.
The following Figure shows the RAP general architecture.

click herefor a bigger image.

The project RAP was started in April 1994 under grant ITW 9404 0 by the German Ministry for Education, Science, Research and Technology (BMBF) for a period of three years and continues work done in PHI.
Further Information: Related Projects:
[DFKI][lab][up][search][server info]


contact: biundo@dfki.uni-sb.de last change: 01/15/97