EEG-based action anticipation in human-robot interaction: a comparative pilot study.

Area of Science:

• Robotics and Neuroscience
• Artificial Intelligence and Human-Computer Interaction

Background:

• Robots are increasingly integrated into various industries, necessitating improved human-robot collaboration.
• Anticipating human actions is key to enhancing safety and efficiency in human-robot interaction (HRI).
• Electroencephalographic (EEG) signals offer potential for predictive capabilities due to their ability to detect pre-movement brain activity.

Purpose of the Study:

• To explore the use of EEG signals for action anticipation in HRI.
• To leverage the high temporal resolution of EEG and modern deep learning techniques for predictive robotics.

Main Methods:

• Evaluation of multiple Deep Learning classification models on a motor imagery (MI) dataset.
• Utilizing EEG signals to capture brain activity preceding voluntary movements.
• Validation of model performance in a pilot HRI experiment.

Main Results:

• Achieved up to 80.90% accuracy in classifying motor imagery tasks using Deep Learning models.
• Successfully predicted human actions several hundred milliseconds prior to execution in a pilot study.
• Demonstrated the feasibility of EEG-based action anticipation in a real-world HRI context.

Conclusions:

• Combining EEG signals with deep learning models significantly enhances the potential for real-time action anticipation in HRI.
• This approach can lead to safer and more efficient collaborative tasks between humans and robots.
• The findings pave the way for more intuitive and responsive robotic systems in collaborative environments.