CONSTRUCTIVE STOCHASTIC TEMPORAL REASONING IN SITUATION ASSESSMENT

Situation assessment tasks (SA-tasks) are important, in many applicati ons, dealing with real-time environment. Due to poor structure of thes e tasks, the knowledge-based approach has been studied intensively by many authors. Regrettably, little progress up to now has been achieved because the existing techniques usually neglect the constructive natu re of a human expert's reasoning in a process of a SA-task solving. Th e latter feature manifests itself in the unconscious creation by an ex pert of new temporal/spatial patterns which, in a general case, are no t present in an explicit form in the processed evidence data. Due to t hat, the development of a new method enabling one to model constructiv e reasoning, is rather complicated, the scope of this work has been re stricted to the temporal aspect only. In the paper, a mathematical fra mework, which may be used in a rule-based expert system, reasoning abo ut events in a constructive manner, is presented. Time instances and i ntervals are treated in detail, using a stochastic approach to represe nt imprecision. The human expert's temporal knowledge is considered im precise in the stochastical sense, as well. Both forward and backward modes of reasoning are studied. Sufficient conditions for consistency of the knowledge model are also derived. The method has been implement ed by the author in a feasibility demonstration prototype expert syste m.