Rewards-driven control of robot arm by decoding EEG signals

Area of Science:

• Neuroscience
• Robotics
• Machine Learning

Background:

• Decoding user intentions from electroencephalography (EEG) signals is crucial for brain-computer interfaces (BCIs).
• Controlling robotic systems with non-invasive EEG for self-paced movements presents significant challenges.

Purpose of the Study:

• To investigate the feasibility of predicting robot arm control goals from non-invasive EEG signals during self-paced reaching movements.
• To develop and evaluate a system for online goal estimation and optimal trajectory generation for robot arm control.

Main Methods:

• Utilized online classification of EEG signals to continuously estimate the movement goal.
• Generated optimal trajectories for a simulated 7 degrees of freedom KUKA robot arm.
• Conducted experiments with a healthy subject performing a planar center-out reaching task.

Main Results:

• The proposed system demonstrated the ability to estimate movement goals from EEG signals prior to movement onset.
• Achieved approximately 80% accuracy in reaching the intended goal with smooth trajectory generation.
• Successfully controlled a simulated KUKA robot arm based on decoded user intentions.

Conclusions:

• Predicting robot arm control goals from non-invasive EEG signals is feasible for self-paced movements.
• The developed online estimation and trajectory generation system shows promise for intuitive robotic control.
• Further research can expand this approach to more complex tasks and diverse user populations.