Mobile Robot Motion Research By BCI Motor Imagery

Brian-Computer Interface (BCI) provides a way to exchange information with theoutside world without the participation of nervous system and muscular tissue, but onlythrough the human brain, which makes it possible for limb dyskinesia patients to interactwith the outside world. BCI has broad application prospects in rehabilitation engineering,health devices, entertainment and brain cognition, etc.A Brain-Robot Interface based on motor imagery (MI) is proposed in this paper. Afterusers finish learning to control their thinking, the system makes pattern recognition on thecollected users’EEG signals and finallytranslates them into control commands to control themovement of the robot. The main work of the paper includes:(1) The research status of the Brain-Robot Interface is summarized. The mechanism andpropagation mechanism of EEG and the fundamental of motor imagery signals are introduced.(2) A feature extraction algorithm (Wavelet-CSP) and three distinct classifiers arestudied. The results show that SVM has better classification accuracy than others, which isfinally used in this paper. Meanwhile, a semi-supervised leaning algorithm is proposed toimprove the accuracy of training stage.(3) The theory of robot motion is described. Two different indoor navigation modes-autonomous mode and cooperative mode are proposed. Being used in the system experiment,the cooperative mode can effectively relieve users’ fatigue caused by long time usage.Meanwhile, with the usage of state machine, two kinds of MI are applied in multi-statemovement of the robot.(4) A complete motor imagery EEG robot motion control system is designed anddeveloped. The standardized experiment procedure is designed. The online and offline BCIexperiments are designed and carried out. The robot can be satisfactorily controlled onlineby user’s right and left hand motor imagery signals.Experiments indicate that the system is able to extract users’ EEG signal featurescorrectly, translate them into robot’s control instructions, which can be used to make real-time control on robots effectively.