Understanding Metrics of Vehicle Control Take-Over Requests in Simulated Automated Vehicles

Abstract

This paper derives an effective take-over alarm method for a Society of Automotive Engineers Level 3 automated vehicle using vehicle parameters. An experiment was conducted using a driving simulator. Forty-one subjects participated in two different scenarios: unplanned and planned operational design domain. Subjects drove on a highway at 100 km/h in automated mode. Experiment participants resumed manual control of the vehicle when the alarm was provided in the take-over state. A visual and auditory combined alarm was the most effective in terms of drivers’ cognitive load showing low steering angle deviation and steering reversal rate values compared to the combined visual, auditory, and haptic alarm (p = 0.01 and 0.003). The auditory alarm was the most effective in terms of the quality of the driver’s reaction showing a low maximum acceleration value compared to the combined visual, haptic alarm (p = 0.002). The combined visual, auditory, and haptic alarm was the most effective in terms of post-take-over lateral control showing a low standard deviation lateral position value compared to the haptic alarm (p < 0.001). The auditory and haptic combined alarm was the most effective in terms of post-take-over longitudinal control showing a low number of gas pedal inputs value compared to the haptic alarm (p = 0.002).