An Asynchronous Mi-Based BCI for Brain-Actuated Robot Grasping Control

Abstract

The brain-actuated robot grasping control is a hard and complex problem due to the low classification accuracy of mental pattern and the high degree of freedom of robot hand and arm. We propose an asynchronous MI-based BCI that allows the user to control the moving direction of robot hand and arm by intension whenever he feels necessary to adjust the moving path. In order to improve the classification accuracy and real-time performance, the thresholding method and LDS modeling approach are employed for online motor imagery pattern recognition. Then, the shared control strategy combines the automatic grasping control of robot and mental control by our asynchronous MI-based BCI to grasp the object with obstacle. Finally, the two experiments illustrate the advantages of our methods in the classification accuracy and grasping performance aspects.