Examination of Operator State Monitoring and Operator Engagement as Strategies for Mitigating Human Factors Challenges Associated with Transfer-of-Control During Automated Driving


Anuj Pradhan, PhD
Assistant Research Scientist, Human Factors Group, U-M Transportation Research Institute


May 2015 – April 2017


Shan Bao, PhD
Assistant Research Scientist, Human Factors Group, U-M Transportation Research Institute

C. Raymond Bingham, PhD
Research Professor, Psychiatry, School of Medicine; Health Behavior and Health Education, School of Public Health; and head of the Young Driver Behavior and Injury Prevention Group, U-M Transportation Research Institute

John Sullivan, PhD
Associate Research Scientist, Head, Human Factors Group, U-M Transportation Research Institute


This project has two broad objectives: To understand the role of in-vehicle operator state monitoring in the transfer of control between human operator and vehicle automation; and to examine driver engagement strategies and their effects on out-of-the-loop operators during automated driving.


A series of experimental studies will be conducted on an automated vehicle driving simulator with integrated driver monitoring capabilities to measure and analyze operator response to transfer of control at (NHTSA) Level 2 & 3 automation of (a) operator monitoring, and (b) operator engagement systems.


Findings inform Mcity’s understanding of the challenges and opportunities of deploying connected and automated vehicles in the following ways:

  • The human factors challenges of the operator’s role in the transition between manual and automated driving is well recognized. This study’s findings help inform strategies to mitigate such issues.
  • Keeping drivers engaged in the driving-loop during lower levels of automated driving (up to Level 3) requires real-time information about the operator’s state so that he/she can be re-engaged appropriately.
  • Drivers’ visual gaze behavior can be monitored to sense driver’s attentional states and can be used for helping drivers better engage with the driving task.
  • The current method uses a parsimonious approach leveraging existing and non-invasive sensors. Future work can involve supplementing with additional sensors for more signals and increased robustness.