PEGASE+ is an engine dedicated to program
supervision knowledge-based systems, which can automate the choice and execution
of programs from a library to accomplish a processing objective. The programs
and their use are now considered as a part of the companies
patrimony that should not be lost and that should be easy to re-use
and maintain. For this purpose, companies need to ``capitalise''
in a formal and readable form the necessary skills for the optimal
use of programs, for documentation purposes and knowledge management in
the companies.Techniques of program supervision have a twofold objective:
both to favour the capitalisation of knowledge about the use of complex
programs and to operationalise this utilisation for users not specialised
in the domain.
Instead of developing each system from scratch, we design engines, independent rom specific applications, but yet dedicated to the particular task of program supervision. A program supervision environment, provided by a software platform (named Lama) is used by a specialist of a library of programs to build a knowledge base. The result (obtained by the integration of the engine plus the knowledge base plus the library of programs) is a knowledge-based system for program supervision which can be utilised by an end-user to run an application.
The Pegase+ engine provides a HTN (hierarchical-task network) planner,
an execution module, some evaluation facilities and a repair mechanism
using repair and adjustment criteria. It works as a trial and error mechanism
until the results are correct.
Pegase+ [IEE00, RFIA00]
has been developped using the Lama software platform. It is written in C++
with a corresponding (optional) graphic interface written in Java.
Pegase+ is currently our most developed and tested engine in program supervision.
Since 1994, several knowledge bases have been developed with Pegase+ in different
application domains [ICVS99].
For example, in image processing for automatic target recognition [IVC99],
in medical imaging [SPIE97], or
for satellites images [IJRS96].
Given a user's request and a knowledge base, the engine develops an
execution plan\index{plan}, which correspond to the successive expansions
of composite operators and execution of primitive ones. In fact it develops
several tentative plans, some of them may be aborted, due to problems detected
during reasoning. The solution plan, if it exists, is produced as a result
of program supervision, as well as the output data.
Contact Sabine.Moisan@sophia.inria.fr
Projet Orion
INRIA
2004 route des Lucioles - BP 93
06902 Sophia Antipolis cedex
France