Balancing an Inverted Pendulum with an EEG-Based BCI

Abstract

To research the brain computer interface (BCI) for dynamic objects control, in this study, we constructed a BCI paradigm for balancing a virtual inverted pendulum on a cart (IPC). In the paradigm, subjects balanced the pendulum by imaging left/right movements. Not only was the direction, but also the strength of motor imagery was estimated simultaneously from the EEG signals, to generate suitable control force for IPC. Additionally, to solve the inconsistent problem between offline training and online controlling, a special online training experiment was designed to obtain more robust parameters of BCI. Three graduate subjects participated in this study, and two of them fast grasped the skill of IPC balancing, achieved balancing time of about 20 seconds. The results showed that the paradigm in this study was feasible and efficient for dynamic objects control.