Incremental processing of temporal observations in Model-Based Reasoning

Observations play a major role in Model-Based Reasoning. In an uncertain, event-driven perspective, the observation of a dynamical system over a time interval is not perceived as a totally-ordered sequence of observable labels but, rather, as a directed acyclic graph. Problem solving, however, requires generating a surrogate of such a graph, the index space. In addition, when tasks such as monitoring and diagnosis are carried out, the observation hypothesized so far has to be integrated at the reception of a new fragment of observation. This translates to the need for computing a new index space every time. Since such a computation is expensive, a naive generation of the index space from scratch at the occurrence of each observation fragment becomes prohibitive in real applications. To cope with this problem, the paper introduces an incremental technique for efficiently modeling and indexing temporal observations.