Identification of Multimodal Pilot Control Behavior in Real Flight

Abstract

Flight simulators are widely used for research and training of pilots. However, in skill-based control tasks, pilots behave differently in the simulator compared to real flight, due to limited visual and physical motion cues. This warrants a method that quantifies simulator fidelity by the differences between pilot control behavior in real flight and in the simulator. This paper presents the results of in-flight experiments to determine multimodal pilot control behavior in real flight, the baseline in determining simulator fidelity. The experiment is performed in a Cessna Citation II laboratory aircraft with a custom-built fly-by-wire system. To estimate pilots' visual and vestibular responses, two forcing functions need to be inserted into the control loop at different locations. The fly-by-wire system was used to physically disturb the aircraft with a disturbance forcing function, while pilots had to track a target forcing function on a display in the cockpit. Both a multisine and a ramp target signal were tested in roll and pitch control tasks, resulting in four experimental conditions. The results show that multimodal pilot control behavior can indeed be identified in real flight using the current fly-by-wire system setup. For two pilots, pilot model parameters could be estimated with high accuracy for all conditions. For some conditions of the two remaining pilots, however, no accurate parameter estimates could be found, as their control activity did not allow for a stable global optimum of the parameter estimation problem.