Modeling Pilot Control Behaviorfor Flight Simulator Design and Assessment

Abstract

Linear control theoretic models of human pilot behaviour have existed for approximately fortyve years. These models have proven useful for understanding aircraft handling qualities. More recently a number of these pilot models have been used to study the e ects of ight simulator delity on the pilot/aircraft control behaviour. To be useful for this type of analysis, the pilot model must incorporate at least a rudimentary model of human motion perception. Typical simulator users are not currently aware of the usefulness of these models for this application. With this objective in mind, four of the more common pilot models: the Crossover, Precision, Structural and Descriptive models are described in detail. These models are then t to a single degree-of-freedom yaw disturbance rejection task in an Apache AH-64 helicopter, and in a xed-based simulation of the same task/helicopter. In addition, the Structural, and Descriptive models, which include rudimentary models of the human visual and gravito-inertial systems, are t to a moving-based simulation of the same task/helicopter. All of the models predict the general trends of reduced performance due to the lack of motion cues and simulator visual time delays. The Structural and Descriptive models also predict the general trends of increased performance when motion is added to a xed-base simulator; however, the magnitude of performance increase was over-estimated in comparison to simulator experimental data. In addition, the models predict a di erent mechanism for the improvement in performance compared to the simulator experimental data. The models are therefore useful as a general analysis tool for simulator users, but they cannot currently predict the exact impact of simulator cue imperfections on pilot behaviour in novel situations.