Trust State Monitoring of Human-machine Co-driving Based on EEG Signals in Intelligent Driving Scenarios

The rapid development and widespread adaption of Artificial Intelligence (AI) technology are reshaping the dynamics of human-machine interaction. Intelligent driving, as a prominent representative of the latest technological revolution, achieves precise control over driving tasks through collaboration between the driver and AT systems. In this context, the driver's trust in AI plays a pivotal role in influencing the efficiency of human-machine cooperation and overall driving safety in intelligent driving scenarios. Nonetheless, there remains a notable absence of methods for accurately assessing and monitoring the state of human-machine interaction, particularly with regard to trust levels during co-driving. To address this gap, this study established an experimental platform utilizing UC-win/road software and conducted experiments to capture and analyze the driver's EEG signals. The primary objective was to develop a neural network-based trust discrimination model. This model exhibits a remarkable level of accuracy, achieving a data analysis accuracy rate of 90%. Furthermore, the research findings indicate that electrodes positioned in the occipital region of the brain exhibit heightened sensitivity to variations in human-machine trust levels compared to other electrodes. Notably, a trust discrimination model constructed solely using these occipital electrodes achieved an impressive accuracy rate of 98.52%. The significance of this trust discrimination model lies in its potential to dynamically measure human-machine trust levels, enabling real-time predictions and early warnings regarding AI interaction risks in intelligent driving scenarios.