Development and Evaluation of Trainee Performance Measures in an Automated Instrument Flight Maneuvers Trainer.

摘要

A simulator study was conducted to improve training performance measurement selection methods, apply the results to an automated flight training system and conduct an evaluation of resulting measurement during automated training of four instrument flight maneuvers. Empirical methods were used to select from an analytically derived set, those measures which had the ability to discriminate between early and later training performance. The multiple discriminant model emerged as the best technique, but the algorithm for its use was highly modified. The automated trainer was then modified to operate on three measurement subsystems, (1) the original scoring algorithm, (2) the measures and weighting coefficients based on multiple discriminant analysis results, and (3) the original scoring algorithm using measured normative data. Resulting measurement was evaluated by automatically trained three matched groups of five civilian pilots each with the result that time-to-train was reduced 34-40% for pilots training with empirically derived measures over the original scoring algorithm. It was recommended that data collection at an operational site be undertaken to verify the methods and to produce information that might lead to a measurement specification for future devices. Recommendations concerning the design of adaptive logics were made. (Author)