Model-based diagnosis and debugging

Model-based Reasoning was originally based on the realization that the description of physical devices and systems in declarative fashion can dramatically facilitate the flexibility of knowledge base construction when reasoning about the systems' behavior and function. Traditionally the main application for this has been in Model-Based Diagnosis where generic reasoning algorithms are applied to models (i.e., formal declarative descriptions) of basic components from a particular domain to identify and rank faults in complex systems composed from these components, in the presence of incomplete information and multiple simultaneous faults. Extensions of the paradigm have covered testing, repair and reconfiguration, and for several years also the application of model-based diagnosis to software artifacts. This area, model-based debugging, poses challenges both in the different kinds of models required and in the difficulties posed by finding diagnoses in a design environment. The talk will present the genesis of Model-Based Debugging from its diagnosis origins and the state of the art in MBD methods.