Content area
Accurate prediction of systemic human cognitive states is critical for developing robust autonomous Human-Robot Interaction (HRI) teams in which artificial agents and human teammates work together for a shared goal. These artificial agents have to operate in a way that does not unnecessarily increase human teammates’ cognitive load as well as adapt their behavior and the teams’ task allocation proper to human teammates’ cognitive states to attain a better work balance which potentially enhance the effectiveness of the whole team. To achieve these, different physiological signal modalities can be used as an objective measurement to assess multiple systemic human cognitive states including workload, sense of urgency, mind wandering, interference, and distraction. Yet, it is currently still unclear which sensing type might allow artificial agents to derive the best evidence of human cognitive states. In Part 1 of this dissertation, we examined the efficacy of various physiological biomarkers, obtained from different signal types such as eye gaze, Electroencephalogram (EEG), Arterial blood pressure (ABP), Functional near-infrared spectroscopy (fNIRS), skin conductance, and respiration, in predicting these human cognitive states in simulated driving scenario. Our results demonstrated that eye gaze has the most superior performance in predicting multiple human cognitive states including workload, sense of urgency, and mind wandering. Then in Part 2, we designed an online workload estimation paradigm within a simulated space station platform and established an HRI scenario involving two robots and two individuals. In this setup, the robots dynamically adjusted their behaviors based on real-time assessments of human cognitive workload. This part of the study focused on evaluating how such adaptive robotic behaviors influenced team efficiency and task performance.