An Efficient Problem Solving Architecture Using ATMS

Subtitle: Tested on a Non-Toy Case Study

Article type: Research Article

Authors: Arlabosse, Francois | Jean-Bart, Bruno | Porte, Nathalie | de Ravinel, Beatrice

Affiliations: Research and Development Dept., Framentec S.A, Tour Fiat, cedex 16, 92084 Paris la Defense, France, uucp:farlabos,bjeanbart,nporte,bravinel@framentec.fr

Abstract: An overall goal for second generation KBS is to develop a theory of reasoning based on the structure and behavior of a physical system. To understand how a physical system works is essential for diagnosing it. Conversely, diagnosing a device is a good territory to test our theory of deep reasoning. In order to solve a diagnostic problem, we have developed and tested a first case study in the Esprit P820 project. It is an inference engine using the ATMS (Assumption-based Truth Maintenance System) (de Kleer, 1986) capable of formal reasoning, modeled on the applications of the qualitative program described by de Kleer and Brown and applying, in an original way, the causal ordering system proposed by Iwasaki and Simon. The concrete realization of this inference engine - that we will redesign and test in another case study of industrial size - will permit a better understanding of qualitative reasoning and allow us to make progress toward a diagnostic theory of a physical system, based on the first principles approach.

DOI: 10.3233/AIC-1988-1402

Journal: AI Communications, vol. 1, no. 4, pp. 6-15, 1988