Augmenting robot behaviors using physiological measures of workload state

Abstract

The evolution of robots from tools to teammates requires a paradigm shift. Robot teammates need to interpret naturalistic forms of human communication and sense implicit, but important cues that reflect the human teammate’s psychological state. A closed-loop system where the robot teammate detects the human teammate’s workload state would enable the robot to select appropriate aiding behaviors to support its human teammate. Physiological measures are suitable for assessment of workload in adaptive systems because they allow continuous assessment and do not require overt responses which disrupt tasks. Given the large variability in physiological workload responses across individuals, an algorithm that accommodates variability in workload responses would be more robust. This study outlines the development and validation of algorithms for workload classification. It discusses (i) a workload manipulation paradigm, (ii) the evaluation of the algorithms for deriving a workload index that is individualized, and (iii) parameter selection for optimal classification.