An efficient algorithm for temporal abduction

Abduction is the process of generating explanations for an observation, starting from a domain theory. We consider an extension of the classical definition of abduction in order to deal with temporal knowledge both in the theory and in the observations. This form of abduction is the basis of many problem solving activities such as temporal diagnosis and reasoning about actions and events. We show how the locality property can be used in order to compute temporal explanations efficiently. In fact, they could be computed using first an abductive reasoner as a generator of candidates and then a temporal reasoner for checking the consistency of each candidate.