Configuration as diagnosis: Generating configurations with Conflict-Directed A^∗ - An application to training plan generation -

Although many approaches to knowledge-based configuration have been developed, the generation of optimal configurations is still an open issue. This paper describes work that addresses this problem in a general way by exploiting an analogy between configuration and diagnosis. Based on a problem representation consisting of a set of ranked goals and a catalog of components, which can contribute in combination to their satisfaction, configuration is formulated as a finite constraintsatisfactionproblem. Configuration is then solved by state-search, in which a problem solver selects components to be included in an appropriate configuration. A variant of Conflict-Directed A^∗ has been implemented to generate optimal configurations. To demonstrate its feasibility, the concept was applied, among other domains, to personalized automatic training plan generation for fitness studios.