A Predictive Model of Driver Response in an Autonomous Environment

Abstract

Remarkable strides have been made in the area of driving safety. Advanced technologies have successfully mitigated some crashes and there is a renewed belief that automation may be the solution to at least some of the remaining safety issues, particularly ones centered on human errors. Whilst automation is expected to be able to execute most longitudinal and lateral control tasks (Gold & Bengler, 2014), we do not yet know how effectively an automated driving suite (ADS) can re-integrate the human within the loop, especially in those situations where a driver’s response must be initiated quickly and is safety-critical. Successful transfer of control is made more complicated by various driver states (e.g., driver trust in automation, quality of calibration, and level of skill) and situational variables (e.g., sudden failure of the automation) that influence seamless transfer. The conceptual model proposed here, referred to as the AC/DC (Automatic Control/Driver Control) model and grounded in constructs of trust, calibration and ability, can be used to predict how a driver will respond in low and high risk scenarios when the automation fails suddenly and without warning. Such a model has the potential to expand our understanding of the complex interactions among different driver behavioral and situational factors that lead to particular responses and could inform interface design or human- automation control transfer protocols.