Design and Experimental Validation of Brain-Computer Shared Control of a Robotic Arm Based on Motion Intention Prediction

As one of the commonly used signals in brain-computer interfaces, direct control of a robotic arm by motor imagery signals introduces ambiguity to the patient's intent recognition. In order to disambiguate and predict the patient's motion intention, this paper uses a fuzzy Petri net to fuse disambiguation and prediction metrics to dynamically regulate and assign weights to the control system to achieve accurate recognition of motion intention. Firstly, a prediction metric based on short-term directional deviation and short-term path distance is introduced to solve the problem that the two target objects to be determined are at similar distance and in a straight line with the end of the robot arm, respectively. Secondly, the disambiguation metric is introduced to improve the accuracy of target prediction, to achieve accurate assisted grasping tasks with low classification EEG signals, and to reduce the user burden in the shared control system. Finally, the effectiveness of this paper's algorithm on intention disambiguation and prediction is experimentally verified.