Closed-Loop Extracranial Activation using Reinforcement-learning (CLEAR)

Increased workloads and operational pressures can degrade human analysts cognitive performance, jeopardizing their ability to safely and effectively carry out mission-critical tasks. To avoid overload and maximize the potential of human operators, a method for conducting real-time evaluation of cognitive state, combined with means to dynamically enhance performance, is required. Novel technologies for closed-loop feedback control of non-invasive brain stimulation can provide meaningful assessment, analysis, and augmentation of dynamic brain information processing capacities (dBIPC). Assessments of electrical and hemodynamic brain activity, combined with available behavioral measures, can provide the information necessary to evaluate state and optimize performance using non-invasive brain stimulation. To improve Warfighter performance on mission-critical tasks, Charles River Analytics proposes to design and demonstrate a system for Closed-Loop Extracranial Activation using Reinforcement-learning (CLEAR), a hardware agnostic, closed-loop system that monitors, detects, and safely manages individual stimulation parameters using reinforcement learning with flexible reward and policy mechanisms. CLEAR combines real-time electroencephalography (EEG) and functional near-infrared spectroscopy (fNIRS) signal processing to unobtrusively assess and ultimately predict cognitive state. CLEAR then uses reinforcement learning techniques to optimize stimulation parameters to deliver precision targeted modulation to select brain regions and enhance performance for extended periods of time.