Framework for the BCI-Motor Imagery Based Control of Humanoid Robot

This paper describes a Brain-Computer Interface (BCI) workflow using the ROSNeuro pipeline and recorded EEG data. The primary objective is to showcase a fully integrated process—from data acquisition and recording to signal processing, classification, and final command output, and interfaces them with a virtual AlterEgo robot in the Gazebo simulation environment. A Gaussian classifier is trained on motor imagery (MI) data to distinguish between two classes. Predictions are translated into actuator commands using standardized ROS message formats for a navigation and manipulation scenario. A comprehensive diagram illustrates how classified MI signals within a finite state machine (FSM) interface allow users to navigate menus and select robot tasks despite limited input commands. The system demonstrates how the AlterEgo robot can effectively augment the capabilities of neuro-motor impaired subjects, leading to autonomous operation in various scenarios.