Research On Electroencephalogram-based Control Methodologies For Exoskeleton Robot

With the largest population in the world,China is faced with the difficult challenge of population aging.Population aging is an important cause of the rapid increase of the disabled and elderly population in China.Considering the increasing demand for assistance or rehabilitation of these people,it is imperative to promote the research and application of assistive technologies for the disabled and the elderly.As the continuous development of robotics technologies,exoskeleton robot has become an effective tool for assistive and rehabilitative purpose.Generally,The wearer of the traditional exoskeleton is passively pulled by the exoskeleton to perform rehabilitative or assistive actions,which is not in accordance with the inherent rhythm of human movement.In order to improve the performance of the exoskeleton robot system,a lot of research has introduced brain-computer interface system.The brain-computer interface can effectively capture the wearer's action intention,thus allowing the wearer to control the exoskeleton voluntarily to perform rehabilitation or assisted actions.In this way,the effect of rehabilitation and assistance can be improved.However,existing studies are based on traditional paradigms such as evoked potentials or motor imagery,which are not friendly to the wearer and add to the complexity of operation.The purpose of this study is to propose innovative Electroencephalogram(EEG)-based exoskeleton robot control methodologies;within the paradigm based on the proposed control methodologies,the wearer can control exoskeleton robot more easily and conveniently.Firstly,by establishing a mapping relationship between obstacle-surmounting gait data and corresponding EEG data,walking gait and obstacle-surmounting gait are used to mark the EEG data.The marked EEG data are used to train the support vector machine classifier to recognize the subject's intention of stepping over the obstacles in order to realize an obstacle-surmounting paradigm.In this paradigm,when a wearer of the exoskeleton robot encounters an obstacle,he can naturally control the exoskeleton robot with his own EEG signal to step over the obstacle without any additional operations.Secondly,the time for configuring brain electrodes in the obstacle-surmounting paradigm is too long,which leads to the degradation of data quality.In order to solve the problem,this study proposes an optimization strategy for brain electrode channel selection based on a heuristic algorithm.This strategy evaluates each brain electrode based on the apriori common spatial pattern of the EEG signal,and then screens out a subset consisting of the brain electrode channels that are highly relevant to the obstaclesurmounting task.Based on the selected channel subset,the time for brain electrode configuration can be greatly reduced with a tolerable performance loss,thus improving the efficiency and quality of the experiments.Lastly,according to the proposed obstacle-surmounting paradigm and the traditional motor imagery paradigm,the underlying control logics are designed for the selfdeveloped exoskeleton from our research group,so as to conduct a comparative analysis.Compared with the traditional motor imagery paradigm,the proposed control method can simplify the operation complexity of controlling the exoskeleton robot to step over obstacles for the wearer.Proposed in this study,the control methodologies which are based on the mapping relationship between gait data and EEG data can realize a wearer-friendly EEG experimental paradigm,and thus provides a new vision for the field of brain-controlled exoskeleton robots.