Error-Related Potential During Cooperative Exploration by Humans and Robots in a Large-Scale Maze and the Effect of Boredom on It

In future, humans live with autonomous robot, and cooperate with it. In this study an autonomous robot and a human were asked to cooperate in an exploration task to reach a goal in a large-scale maze. While they were doing a task, electroencephalogram (EEG) was measured from the human. The human was asked to choose the direction of travel at intersections, and 30% of the choose, the robot traveled in a different direction from that specified by the human. Error-related potentials (ErrPs) occurred in the human brain at the time, and that was consistent with that reported in the previous studies. Time-frequency analysis of EEG revealed an increase in θ-frequency power and a decrease in $\beta$-frequency power. Some participants felt “bored” during the task. As if to reflect this, in the second half of the task, the amplitude of ErrPs decreased and $\theta$-frequency power also decreased. When the ErrPs of the first and second halves were discriminated using a convolutional neural network, the accuracy rate of the correct responses decreased in the second half compared to the first half. These results suggest that ErrP changes with boredom, and there is a possibility to detect the boredom using EEG.