Brain Driving: Personalizing Vehicle Speed With DR-EEG Decoding and Situational Embeddings

Driving is a complex and personalized endeavor that entails the brain processing sensory information amidst continuously changing situations. Here, we sought to explore the potential of an electroencephalogram (EEG)-based decision-making approach for predicting personalized driving behavior. Driving behavior data and EEG signals were collected simultaneously from 133 participants. Our analysis revealed that the time-frequency cross mutual information (TFCMI) serves as an effective driving-related EEG (DR-EEG) feature capable of describing the correlation between brain regions irrespective of their physical proximity. Furthermore, we demonstrated that incorporating situational embeddings yielded a significant improvement in predictive performance of 27.28%. Notably, our results highlighted the impact of individualized differences in brain cognition and situational perception, which enable individual models to outperform general models, with an average of R2 = 0.982. We found that the activation of neural connections associated with the frontal and left temporal lobes appears to be of paramount importance in driving decision-making. These findings suggest that integrating individual brain and situational knowledge into the computational decision-making framework of driving holds immense promise for deploying driving-related brain-computer interactions (BCIs) and developing personalized intelligent driving systems.