LIP6 1999/003

• Habilitation «Méthodologie de conception des Systèmes d'Aide à l'Exploitation des Simulateurs d'Entraînement»
• M. Joab
• 119 pages - 01/12/1999 - document en - http://www.lip6.fr/lip6/reports/1999/lip6.1999.003.ps.tar.gz 371 Ko
• Contact Michelle.Joab (at) nulllip6.fr
• Ancien Thème : SYSDEF
This report concerns man-machine interaction and knowledge engineering. The first part of this report describes a methodology for designing operating support systems of training simulator (OSSTS). The OSSTS is devoted to vocational training by the means of a simulator. The concerned operators are trained in order to control and supervise dynamic devices. Since many years, simulators were used as training support for the operators according to the paradigm of learning by doing. A OSSTS is a Training Simulator enriched by sophisticated help functions, for the instructor and the trainee. A OSSTS design holds currently the difficulties of an Intelligent Tutoring System and those due to the technical complexity of the learning field. Starting from considering industrial projects, an original design methodology which performs knowledge capitalization within OSSTS has been elaborated. Capitalization means reusing knowledge, deducing new knowledge or collecting new knowledge on the basis of already gathered knowledge. For the "tricky" components of the OSSTS needing the instructor expertise, which is often unaffordable, we suggest to build a bootstrap system in an iterative development cycle . With this aim, we suggest to reuse generic reasoning models or to drift reasoning models for other models of the OSSTS. Knowledge acquisition for the DIAPASON system illustrates our methodological approach. DIAPASON is to be used by the operators in charge in the telecontrol of middle voltage power systems. This synthesis study clarifies knowledge models. The second part of the paper deals with our studies on modeling explanatory dialogues. We suggest a characterization of explanatory sequences based on the recognition of the structural properties of the dialogue and the illocutionary function of speech acts. We validate our approach by corpus studying. To manage the dialogue, we have chosen a postponed structuring mechanism rather than a predictive structuring. A mock up of the dialogue system has been implemented to validate the characterization of the explanatory sequences. We conclude this report with two research directions : to enlarge the individual training to team training taking advantage of distributed simulators, specify architecture patterns for the OSSTS resulting from different interaction styles.