| dc.description.abstract | Changing the ways that robots interact with humans in uncertain, dynamic, and high-intensity environments (e.g., a NASA control room) is needed in order to realize effective human-robot teams. Dynamic domains require innovative human-robot teaming methodologies, which are adaptive in nature, due to varying task demands. These methodologies require mechanisms that can drive the robot's interactions, such that the robot provides valuable contributions to achieving the task, while appropriately interacting with, but not hindering the human. The human's complete workload state can be used to determine robot interactions that may augment team performance, due to the relationship between workload and task performance. This dissertation developed a workload assessment algorithm capable of estimating overall workload and each workload component (e.g., cognitive, auditory, visual, speech, and physical) in order to provide meaningful information to an adaptive system. The developed algorithm estimated overall workload and each workload component accurately using data from two human-robot teaming evaluations: a peer-based and supervisory-based. A non-stationary evaluation was conducted in order to validate the algorithm’s real-time capabilities. The workload assessment algorithm was incorporated into an adaptive human-robot teaming system architecture, which targeted adaptations towards a workload component. A pilot study demonstrated the adaptive system’s ability to improve task performance by adapting system autonomy and interactions. | |
