EEG-Based Brain-Computer Interface for Robotic System Control Using Right-Left Hand Movements with Ensemble Model

In this work, machine learning models are utilized to interpret EEG data for the development of a brain-computer interface (BCI) system that controls a robotic car. Voluntary hand motor actions are employed for interface controls, with each hand corresponding to a different direction. EEG data is recorded using a 14-channel EMOTIV EPOC+ device during motor activities such as moving the car left, right, or stopping. The raw EEG signals are preprocessed using Fast Fourier Transform (FFT) before training a range of models, including Random Forest, XGBoost, CatBoost, Bi-Directional GRU, Transformer-based, and a Stacking Ensemble Technique (SET). The highest accuracy is achieved by SET (87%), followed by XGBoost (82%). These results demonstrate the robustness of ensemble approach SET for EEG-based tasks. Although further advancements in real-time processing are necessary, the system's ability to successfully control the robotic car underscores the potential of BCIs in robotic control and assistive systems development.