Modelisation de taches pour la resolution de problemes en cooperation systeme-utilisateur
Jutta Katharina Willamowski · HAL (Le Centre pour la Communication Scientifique Directe) · 1994
A cooperative problem solving environment should allow the user to intervene and steer the solution process. Control over the level of intervention should also be permitted. For instance, it should be possible for the user to fully steer the solution, or for the system to proceed with minimal or no user intervention. In order so enable such flexible system-user cooperation the reasoning process must be exploitable by the system and must be easily understandable by the user. This thesis uses hierarchical planning to model the problem-solving process. The proposed model relies on the concept of a task. A task enables one to define problems at several levels of abstraction. A problem solving strategy can be associated to with each of these problem definitions. Such a strategy describes the recursive decomposition of a task into more elementary subtasks. This model allows the system to efficiently manage the problem-solving process : hierarchical planning allows decomposition and execution steps to alternate. Cooperation can be initiated at every level of abstraction and decomposition : at each of these levels the user can intervene into the problem-solving process, can take control, or can give it back to the system. Hence, in the course of the solution process the problem solving strategy can change and adapt to the appropriate level of system-user cooperation. This model has been implemented into a generic tool, called SCARP, which allows the representation and exploitation of the complete problem-solving knowledge in co-operation with the user. The tool has been used, tested, and validated in the context of different applications. This thesis presents the design of SCARP and describes its use, with examples from signal processing and data analysis. -- Taille du fichier postscript : 2288 KB