The specific character of real-time decision support systems oriented at dy
namic problem domains and the feasibility of taking into account the factor
of incompleteness and uncertainty in these systems on the basis of the non
-Markov model with learning are considered.