The invention relates to a robot system based on a brain-computer interface and an implementation method. The robot system is mainly characterized in that a brain-computer interface sub-system obtains a brain electrical signal through a measurement electrode and performs feature extraction of the brain electrical signal to obtain the intention of a user; the brain-computer interface sub-system converts the intention of the user into a control command and transmits the control command to a robot sub-system through a communication sub-system; the robot sub-system receives the control command of the brain-computer interface sub-system and controls a robot to move; the robot sub-system transmits video information acquired by self to the brain-computer interface sub-system; and the brain-computer interface sub-system forms a stimulation interface fed back to the user. According to the invention, mobility control of the robot in eight directions through the steady-state visual evoked potential brain-computer interface can be realized; furthermore, the system can control the robot rapidly and precisely without being trained; because the brain-computer interface is free from body movement participation, the user suffering from severe dyskinesia can move to an expected position; and thus, the living quality is increased.