P300 BCI based planning behavior selection network for humanoid robot control

Abstract

We propose planning behavior selection network (PBSN) based brain-computer interface (BCI) for controlling a humanoid robot. BCI provides commands to external devices or computer applications only using user's brain signals. However, few commands from BCI cause user fatigue. PBSN is a hybrid method between reactive system and goal-oriented planning system. PBSN has two beneficial points. One is robustness of reactive system and the other is long-term goal planning of planning system. This only requires high-level commands from the user and frees from make low level command to operate the robot. Finally, it makes possible to reduce user's fatigue. Online accuracy test gives reasonable accuracy rate, and PBSN based online simulation shows possibility as an assistant humanoid robot.