Artificial Intelligence Group Project

ENCODE: ENvironment for COnfiguration DEsign

Copernicus Project 0149 (EU funded 420k ECU)

Partners
Publications
Overview
Results

Project Partners

• The Knowledge Systems Group, in the Knowledge Media Institute, at the Open University, in Milton Keynes, England.
• The Artificial Intelligence Group at the Department of Cybernetics and Artificial Intelligence, Technical University, Kosice

• Tomas Sabol (project leader)
• Marek Hatala
• Marian Mach
• Jan Paralic
• Czech Technical University (Czech Republic)
• DccS, Ltd. (small industrial enterprise, Czech Republic)
• ASEPO, Ltd. (small industrial enterprise, Czech Republic)

Project Overview

ENCODE workbench is based on the VITAL methodology with a special focus on engineering design (ED). It will provide tools for knowledge modelling, visualisation, and implementation of knowledge based systems (KBS). The implementation level is supported by various modules (e.g. forward chaining, backward chaining, constraint satisfaction, frames, etc.). Special attention will be paid to the CLIPS as a target language for the implementation level. ENCODE will make use of PC platform.

ENCODE workbench will support the following basic design steps:

1. conceptual modelling of the selected problem
   The problem and its solution are modelled at the knowledge level by a knowledge engineer. The output is an OCML description of the solution which contains domain description, tasks and control. In accordance with the philosophy introduced in \cite{Newell}, the problem is actually solved at the knowledge level. However, ENCODE also provides a set of mechanisms which could be used for implementing of the tasks.
2. translation of conceptual model to the level of target implementation language(s)
   At this step aim is to implement every primitive (leaf) task in the OCML description by one of the mechanisms provided by the workbench. Following implementation mechanisms will be provided within the ENCODE workbench: backward chaining (BC), forward chaining (FC), constraint satisfaction (CS), and frame language. Moreover, the workbench will make it possible to implement user-defined implementation tools. Each task can be implemented by different implementation mechanism.

Moreover, the ENCODE workbench will support the following additional services:

1. visualisation
   It will be possible to visualise the execution at various levels of abstraction (e.g. the OCML level, FC level, etc.)
2. debugging
   The visualisation modules will be used for KBS debugging. A debugging methodology will be worked out.

Two large pilot applications (initial car design and selection of metal forming technology) will be developed to test the ENCODE methodology and workbench.

Preliminary Results

The first version of the Constraint Satisfaction Engine is available for testing.

Copyright © 1995 Marek Hatala