Following the robot's lead: Predicting human and robot movement from EEG in a motor learning HRI task

A large proportion of human behavior is organized in time in the form of sensorimotor sequences. Learning new behavioral sequences recruits cognitive functions with their neural underpinnings. Here, we characterize how neurophysiological activity revealed in the EEG signal can reflect these behavioral processes. This was investigated in a face-to-face human-robot interaction, where the robot demonstrated a continuous pointing sequence, which the human mimicked. We observed task-related modulation of the event-related spectral perturbations (ERSP) in distinct ways for rest, fixation, and movement sequences. We also observed modulation of the ERSP by the motor sequence learning. Using a Markov-switching linear regression model, we further demonstrated that the EEG signal could be used to decode the human and robot movements. These results are significant both in the context of neural coding of motor performance and learning, as well as in the context of neural coding of joint action, in the face-to-face human-robot interaction.