ADAPTIVE AIRCRAFT CREW TRAINING BASED ON ACCUMULATED EMPIRICAL EXPERIENCE

Under consideration are parametric representations of adaptive flight crew training procedures with the aid of discrete-state continuous-time Markov random processes. It is assumed that each crew has one of the specified attainment grades. Substantial grades of training process correspond to flight maneuvers of relevant difficulties, which are defined by the training manuals. Each maneuver, if it is included at the training schedule, has to be represented by a training process status. The one-to-one correspondence between attainment and maneuver difficulty grades is assumed. Transitions between the statuses in question are determined by the certain rules. Free parameters of the Markov processes are identified to get the best fit of observed and expected frequencies of being in certain system statuses at the given time points. The accumulated empirical experience is represented using observed histograms. Knowing the process status, in which a crew under test turns out to be after performing the last maneuver at a specified time point, and probabilities of being in this status at the specified time point for each attainment grade, it is possible to estimate posterior probabilities of attainment grades. A crew implementing training procedure gets the difficulty grade with the greatest posterior probability. During a training procedure the adaptive principle of maneuver selection is used, according to which difficulties of implemented maneuvers must be in approximate correspondence with the estimations of crews' attainment grades. Significantly smaller numbers of maneuvers that must be carried out by a crew to provide the given attainment grade as well as the higher training efficiency are among the advantages of the approach in use. The results obtained are intended to be a mathematical basis for development of the advanced adaptive training intelligent systems which do not currently exist. The principal application area is diagnostics resulted from work on contemporary Full Flight Simulators. The main purpose of applying the presented approach is to replace the subjective instructor assessment with an objective assessment with the aid of the training intelligent systems.