Online brain-computer interface controlling robotic exoskeleton for gait rehabilitation

Robotic exoskeletons for physical rehabilitation have been utilized for retraining the people suffering paraplegia and enhancing neural plasticity in recent years. However, the subjects are not voluntarily involved in most systems. This work aims to develop a locomotion trainer, which can be controlled by the voluntary motion intention of subjects. A hierarchical framework of three levels is proposed according to the control strategy of the natural nervous and motor system of human. In the high level, the brain-computer interface (BCI) based on steady-state visual evoked potential (SSVEP) is used to recognize four types of intention related to walking modes, that are stop, normal walk, acceleration, and deceleration. The middle level aims to mimic the neural control mechanism in spinal cord. Thus central pattern generator (CPG) is established to generate the smooth walking control patterns. It can adaptively realize four walking modes and switch between different modes according the commands from BCI. The low level of the system is a custom-made leg exoskeleton prototype, which has four degrees of freedom and can generate assistive torque for hip and knee joints of two legs. Six healthy subjects took an online experiment to test the performance of the whole system. The recognition rate to accomplish corresponding mode and switch is above 90% for BCI, and the time delay is about 1.5~2 seconds. The results show that all the subjects can generate their desired walking patterns, and the robotic exoskeleton system can realize satisfactory online control for gait rehabilitation.