Development of a Monitoring System for Driver Readiness in Prolonged Automated Driving

Vehicle automation technology is being designed to handle driving tasks for human drivers. However, this technology is not expected to handle all possible driving conditions successfully in the foreseeable future. The system can fail anytime and may require drivers to take over control within a short period of time. Additionally, automation can induce boredom, daydreaming, and drowsiness due to driver inactivity and can worsen driver readiness to take over control of the vehicle. Driver readiness can be measured using their postural data, gaze behaviors, and emotional expressions. Specific thresholds for these measures can be used to alert drivers to be ready to take over from automated driving or avert them from driving, when necessary. This proposal aims to develop a driver readiness monitoring system to improve their takeover performance using a driving simulator study. The objectives are to identify effective measures to define driver readiness and assess the association between divers’ takeover performance and their readiness measures. Researchers will measure (i) driver readiness using postural data, gaze and head orientations, and emotional status from video recordings analyzed with Face Reader software; (ii) drivers’ categorical subjective responses on readiness; and (iii) takeover performance using reaction time, collision rate, and driving behavior. Different machine learning algorithms will be applied for (i) feature extraction, (ii) feature selection, and (iii) developing classification and prediction models for readiness monitoring. The rationales for this research are to: (a) inform policy makers about an effective driver-readiness monitoring system for prolonged automated driving; (b) provide safer driving conditions and address equity during prolonged driving for older adults, and occupational drivers (Uber, taxi, or city transportation) driving long hours; and (c) enhance educational and research infrastructures combining human-computer interactions and machine learning. Stakeholder involvement from the city, state, and automobile manufacturer will strengthen our tech-transfer and project outcome dissemination.