Detecting ErrPs Signals in HRI Tasks

In the last two decades, electroencephalography (EEG) signals have been used as a relevant source of information in human-robot interaction (HRI). In particular, in the last years Error Related Potentials (ErrPs) have been introduced. These potentials can be leveraged during interaction tasks to mark the mismatch between a robot’s behavior and human expectations. These signals are used to better adapt the robot to human needs, through a control based on these signals. This work aims to investigate ErrPs to study their potential through an experiment, in order to use them as feedback for adapting and correcting a robot system. We present a setup and experimental protocol: the experiment is divided into five tasks with seven subjects. For every task, we have 120 events, with a 25%–35% probability of error. We used Matlab2023a and the toolbox EEGLAB2023.0 for EEG analysis. We performed this experiment with a Baxter robot and the interaction with the robot was done in two different ways, with a keyboard or in a teleoperation scheme. The tasks are designed to reproduce, for example, a problem teleoperated pick and place in the industry.