A Brain-Controlled Mobile Robot System Integrating Deep Neural Networks and Model Predictive Control

Abstract

This paper presents a brain-controlled mobile robot system integrating Task-Related Component Analysis (TRCA) filtering, deep learning networks, and Model Predictive Control (MPC). By incorporating an MPC predictive controller between the brain-computer interface and the mobile robot, the system can monitor obstacle distances, ensuring the safety of the mobile robot. By combining TRCA spatial filtering with deep learning-based DNN methods, this approach significantly reduces the classification decision time of the brain-computer interface, markedly enhancing the system's real-time performance and control capabilities, thereby achieving a dual guarantee of real-time capability and safety. Experimental results show that our method outperforms traditional methods in terms of real-time performance, accuracy, and control capability. Offline experiments demonstrate that the classification accuracy of the brain-computer interface increased to 91.14±3.26<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">%</sup>, with the average decision time reduced from 0.51 seconds to 0.2 seconds, and the information transfer rate improved from 121.6 bits/minute to 307.18±36.12 bits/minute. Online experiments validated the method's potential for application in high real-time demand systems by comparing the robot's task completion rate, task completion time, and path length.