EEG Signal-Based Human-Robot Interaction System for Anxiety Regulation via Reinforcement Learning

Anxiety is increasingly acknowledged as a significant public health issue worldwide, especially among adults. Although traditional treatments such as medication and psychological counseling are effective to some degree, they do not accommodate all patients and can sometimes produce suboptimal outcomes. This study proposed and developed a personalized emotion support system that employs reinforcement learning to adjust its reward values by monitoring changes in EEG signals before and after subjects watch robotic dance movements. This process emulates emotional regulation to identify and modulate anxiety in adults. The system automatically adjusts the subjects' emotions based on observed EEG signal changes associated with viewing different dance movements. Preliminary test results indicate that the system can effectively modulate subjects' emotions. Currently, the system processes only non-real-time EEG data, which implies certain limitations in application; however, the research demonstrates the potential of integrating robotics technology with brain-computer interfaces for emotional interventions.