Predicting Adaptive Human Control Behavior to Changing Controlled Element Dynamics Based on Statistical Variations in Error and Error Rate

Abstract

This paper presents an analysis in prediction of adaptive manual control behavior to sudden changes in controlled element dynamics. A previously proposed model, the ‘supervisory control algorithm’, describes human adaptive behavior using binary decision moments at specific decision region limits of the error and error rate signal. This model was assessed using a compensatory pitch tracking task with sudden variations in controlled element dynamics. An experiment was conducted with six participants in a fixed-base simulator at Delft University of Technology. During the runs the participants had to perform the tracking task as accurately as possible and had to indicate detected controlled element dynamics transitions by pressing a button. The results indicate that the original decision region limits from the supervisory control algorithm do not apply to the way human operators adapt in the set-up of this experiment. From the button press data only one percent yielded a result in compliance with the algorithm, demonstrating the large discrepancy between these decision regions and the actual human detection limits in this experiment task. A final analysis is performed using a detection method based on deviations in statistical properties of pre-transitional tracking. This method demonstrated simulation results close to realistic values with a detection threshold around four times the standard deviations.