This study involved an investigation of patterns of arousal in a driver’s sympathetic nervous system toward a system that can detect high cognitive workload, high emotional load, and acute stress (surprise or panic) experiences in drivers. Humans are exposed to cognitive and physical stressors in the driving environment that can be exacerbated when drivers are operating under the cognitive effects of incidental emotions or secondary task loads. Each type of stressor can negatively affect the control of complex operational systems, such as a vehicle, in different ways. The body’s sympathetic autonomous nervous system (the “fight or flight” system) triggers several physiological changes in the body, such as perspiratory and cardiac dynamics, in order to prepare the human to defend themselves or run away. Physiological sensor systems can therefore be used to collect data on these physiological changes (such as heart rate variability (HRV; Taelman, Vandeput, Spaopon, & Van Huffel, 2009) and skin conductance response (SCR; Boucsein, 2012; Shi, Ruiz, Taib, Choi, & Chen, 2007; Villarejo, Zapirain, & Zorrilla, 2012). Data processing techniques can be applied to identify patterns in physiological response across multiple variables and time windows that are characteristics of potentially problematic driver mental states. As a first step toward developing an online system for detecting potentially-problematic driver mental states, this study sought to identify patterns in HRV and SCR data associated with each of these states, and to investigate the predictive power of such an algorithm for explaining aspects of driving behavior.
Eighty-eight people (with equal representation from each sex and each of two age groups) participated in a driving simulator study. The simulation scenarios consisted of a simulated rural highway with multiple predetermined segments where secondary task loads were applied. After a physiological baselining procedure and adequate practice with the driving and secondary tasks, participants completed scenarios under each of the six driving conditions: a “Normal” drive with a simple roadway and no secondary tasks, four “Loaded” drives that included driving-related loads (e.g., construction zones) and some additional secondary tasks designed to impose analytical cognitive, emotional, and motoric (e.g., texting task) loads. A final scenario included varied combinations of these loads and a unique surprising “Failure” event (triggered acceleration in the vehicle that mimicked unintended acceleration problems). HRV and SCR were measured throughout the experiment using Zephyr Bioharness3 and Shimmer3 sensor systems, respectively, and numerous third-party software packages were used to process the data and identify noteworthy patterns.
This presentation will report the analytical results of three dependent variables: HRV, SCR-amplitude (i.e., the magnitude of characteristic increases in arousal during sympathetic response), and SCR-frequency (i.e., the frequency of skin conductance responses identified within a predetermined time window). HRV (measured over time windows of a few seconds, as well as of several minutes) and SCR-Frequency were significantly sensitive to changes in loading, and some patterns in dynamics among multiple measures can be used to differentiate the designed mental states, although further testing is underway to refine and validate these findings. In particular, SCR- amplitude may be an effective way to distinguish acute stress events, as this measure shows a significant and substantial change within 2.5 seconds after realization of the unintended acceleration event. In addition to discussing these results, the current and future efforts in the development of an online driver “stressor detector” and mitigation system will be introduced in the presentation as well.