Adaptive function allocation for human-machine systems: A comparative study of triggers, strategies and stabilization
Abstract
An adaptive allocation system is capable of dynamically reallocating functions between different agents. In adaptive allocation, the need for reallocation can be determined using different parameters (trigger mechanisms) such as fluctuation in workload or changes in physiology of the human agent. The adaptation strategy can determine the amount of reallocation and set the new level of the automated agent. The first experiment evaluated the effects of trigger type (performance or heart rate) and adaptation strategy (complete reallocation or partial transformation) on the number of errors made. The presence of adaptive allocation reduced the number of errors. The ANOVA results indicate interaction effect between the trigger type and adaptation strategy. Due to the presence of an interaction effect, interpretations of the main effects of trigger type or adaptation strategy type are difficult. The combination of a partial transformation strategy and heart rate trigger mechanism resulted in substantially more (nearly 114%) errors than other combinations. The errors made in the other combinations were not significantly different. The second experiment monitored the level of automation over time for presence of stabilization. There is no strong evidence to indicate that the time to stabilization is less than 30 minutes. As the time to stabilization is more than 30 minutes, inferences about main effect of trigger type or adaptation strategy could not be made. Based on the contingency table analyses, there appears to be an effect of adaptation strategy on the mean level of automation and the effect of trigger type on the time of onset of stabilization. Based on the post-hoc ANOVA analyses, there appears to be an interaction effect between the trigger types and adaptation strategy type on the number of reallocations and the probability of a change in level of automation being observed.
Degree
Ph.D.
Advisors
Landry, Purdue University.
Subject Area
Industrial engineering
Off-Campus Purdue Users:
To access this dissertation, please log in to our
proxy server.