Tailored Performance-based Adaptive Levels of Automation

Abstract

Adaptive automation may help balance system autonomy with human interaction in supervisory control environments. The present experiment evaluated application of performance-based adaptive automation for an image analysis/decision task in a multiple autonomous vehicle simulation. An asymmetrical algorithm was employed in which the performance threshold differed for the two steps of the adaptive cycle: 1) increasing level of automation (to relieve participant when overloaded) and 2) decreasing level of automation (to re-engage participant when less loaded). Results showed that performance-based adaptation of the autonomy level improved both the speed and accuracy of performance on the image analysis task. Most participants indicated that adaptive automation reduced their cognitive workload and aided situation awareness. Additionally, the results suggest that the asymmetrical algorithm used to implement performance-based adaptation helped keep participants at a lower autonomy level where automation-induced problems are less likely.