Feasibility of detecting potential emergencies in symbiotic human-robot collaboration with a mobile EEG

Manufacturing challenges are driving the move from separated workspaces of either humans or robots towards a close, symbiotic collaboration. Symbiotic Human-Robot Collaboration requires both parties to not only share the same workspace, but to also perform tasks simultaneously. This raises questions of mutual awareness, for which safety is a critical factor. Despite advances regarding safety systems, human sensing abilities combined with the intelligence to anticipate potential emergencies cannot be matched. Subsequently, the human operator remains in a critical role regarding safety in Human-Robot Collaboration However, in a collaborative environment humans are expected to use their hands towards the completion of a task. Therefore, in order to achieve resilience for collaborative tasks, there is a need to have a hands free detection mechanism for unforeseen events. This work investigates a human sensor-based emergency stop interface that reacts once the human operator senses or anticipates a potential emergency. A novel approach is presented on how a mobile electroencephalogram (EEG) can be used to detect potential emergencies in Human-Robot Collaboration. An experiment was conducted with 21 participants, ten assembly tasks and three different kinds of potential emergencies. The potential emergencies included the collaborative robot to drop an assembly workpiece, to crush the assembly piece on the worktable, and to perform a simulated malfunction. The EEG data suggests strong similarities in the patterns between the different types of potential emergencies. High accuracies were be achieved with a Decision Tree Model based on Continuous Wavelet Transform peak counting. To optimize detection time, different detection window sizes were compared. The results showed a promising potential of this approach, which it is not intended to replace current safety systems but to enhance them towards a safer and thus symbiotic Collaboration