Human Supervision of Time Critical Control Systems. Addendum

Abstract : Data collection of real-time operator performance under varying levels of workload and stress conditions has been performed at AFRL/RHCP. The data includes psycho physiological indicators (5 electroencephalogram (EEG) channels, vertical electro-oculogram (VEOG), horizontal electro-oculogram(HEOG), and electrocardiogram (ECG)) from 12 participants who monitored a simulated mission involving several unmanned aerial vehicles (UAVs). Substantial part of research efforts was dedicated to studying and modeling of information flow in brain networks, extraction of patterns from brain signals in real time. A number of algorithms for quantitative analysis of psycho-physiological data have been developed, including approaches based on p-order conic programming, support vector machines, non-parametric statistical analysis via Granger causality, etc. The conducted studies indicate that the methods based on temporal trend detection in dimensionally reduced time series, obtained via projection of Kullback-Leibler divergence, as well as independent component analysis, possess substantial robustness and can serve as predictive metrics for implementation on closed-loop architectures.