Consistency-based problem solving for environmental decision support

Complex systems do not always behave as we would like them to. With the complexity of the system, be it a water treatment plant, an ecological system or a technical device, the tasks of situation assessment (finding out what is the actual state of the system) and therapy recognition (finding out what can be done to influence it in a desirable direction again) require more and more complex reasoning. This paper proposes a general approach to computational support for these tasks, namely consistency-based problem solving. Building upon research in model-based systems and, more specifically, consistency-based diagnosis, we have developed a revision and generalization of traditional (component-oriented) theories and techniques of diagnosis from first principles. Our approach is both more general in terms of the class of problems to be addressed and more specific by proposing and exploiting a structured representation of system and domain knowledge. A motivating example from the domain of water treatment will facilitate the presentation of the theory of consistency-based problem solving and the description of an implemented reasoning system, the Generalized Diagnosis Engine, G⁺DE, that performs this kind of reasoning for a wide range of problem domains.