Identification and Modeling of Driver Multiloop Feedback and Preview Steering Control

Novel (semi-)automated systems are rapidly being introduced into modern road vehicles, but anticipating possibly critical human-machine interaction issues is difficult, because the human driver’s behavior is as of yet still poorly understood. This paper aims to improve our understanding and models of driver steering behavior on winding roads, using Frequency-Response Function (FRF) measurements of drivers’ feedforward, heading feedback, and lateral position feedback dynamics. The steering behavior data were collected in a human-in-the-loop simulator experiment, in which drivers followed the road centerline at constant forward velocity, while being perturbed laterally by wind-gust disturbances. All three measured FRFs can be captured with a multiloop, single preview-point driver model, which has only five parameters. These parameters provide unmatched understanding of – otherwise lumped – driver internal steering processes, quantifying how and what portion of the previewed centerline trajectory is used for control, and how lateral position and heading feedback are weighed. The gained insights may help to reduce driver-automation interaction issues in modern road vehicles, to quantify between-driver steering variations, adaptation and learning, and to design human-like and individualized automatic and shared steering controllers.