ABSTRACT
Vehicle automation will fundamentally change what it means to safely operate a vehicle. Future vehicles will afford the driver opportunities to decrease their active control of the vehicle through taking their hands off the wheel and taking their eyes and mind off the road. The components of the driving task, and thus driver behavior and expectations, will change as the capability of the vehicle to operate under various conditions grows. The automotive industry, regulators, policy makers, and academics alike have recognized the importance of monitoring the state of the driver in this changing landscape. Changes in driver state, in terms of distraction, disengagement, and drowsiness, are all noted in the literature as likely outcomes associated with increasing automation, with the control transitions between the driver and vehicle cited as a primary safety concern. International bodies, including the European Parliament and the National Transportation Safety Board (NTSB), have recognized the need for Driver State Monitoring (DSM). DSM can be based on various methods of input, including vehicle control measures, physiological inputs, and the duration of driving, among others. However, camera-based approaches are recognized to afford the greatest level of specificity in identifying risky behaviors and driver states. In an automated driving environment, a camera-based approach becomes especially critical given that other measures, including some vehicle-based measures, take on diminished significance for driver state estimation, given the reduced interactions the driver has with the steering wheel and pedals. For these reasons this approach to DSM is being actively pursued by almost all automotive original equipment manufacturers (OEMs).
