Examining dependencies between real-world driving behaviors and stress- harnessing CAN-bus and biometric health data for proactive safety management

This study harnesses seemingly unstructured big data and integrates computational and physical components of cyber-physical systems to understand the links between microscopic driving behaviors and real-world driving stress. Using a blend of RGB-D cameras, wearable devices, and CAN-bus loggers, a natural experiment is conducted to collect over 0.2 million temporal samples of real-world driving behavior and health biomarkers for over 180 driving sessions/trips. The dependencies between microscopic driving behavior and stress (captured through biometric markers such as heart rate) are then explored. Descriptive analysis is conducted first to spot meaningful dependencies among different kinematic signals (speed, longitudinal, and lateral acceleration, etc.) and between kinematic and biometric signals (heart rate, blood volume pulse, electrodermal activity, etc.). Given the hierarchical architecture of the data collection process, and the subsequent methodological concern of unobserved heterogeneity, hierarchical multi-level models are estimated to examine correlations between driving stress and driving behavior. Significant between- vehicle and a relatively larger between-trip heterogeneity in driving stress is observed. While a negative correlation is found between longitudinal acceleration/deceleration and driving stress, significant contrasts are also observed in that extreme values of longitudinal/lateral accelerations and heart rates are relatively rare - pointing to potential compensation effects. Empirical results show that speed, longitudinal and lateral acceleration, and other vehicle performance measures are statistically significantly correlated with driving stress and therefore could be used to detect driver stress level. Incorporating trip-level heterogeneity resulted in substantial improvement in goodness of fit. Implications of the findings for designing proactive driver assist and control systems are discussed.