A Fast and Robust Noninvasive Brain-Computer Interface for Online Field Systems Control

Impact Technologies, LLC, in partnership with Brandeis University, proposes to develop a brain-computer interface (BCI) that enables real-time control of autonomous devices in battlefield environments by extracting high-level commands from noninvasive EEG signals from operators. In addition to the standard processes such as artifact removal, feature extraction/selection, and pattern classification, a novel 3-level hierarchical structure featuring supervised offline training, unsupervised online learning, and environment-in-the-loop is proposed. Particularly, an initial BCI will be built with reduced training effort by the offline supervised training module. The performance of the BCI will then be progressively improved by the online reinforcement learning and Bayesian inference functions to accommodate the changes of mental state and environment. The prototype system is based on a good mixture of mature technologies and innovations. Existing in-house facilities will be drawn upon to develop compelling use-case scenarios. The Phase I activities will focus on defining the requirements for the proposed BCI, early design of algorithms, and tuning of the system. In the following phases, Impact will work closely with Brandeis and OEM partners to develop a technology transition plan for military and commercial applications.