Correlation Between Reaction Time, Multi-Modal Feedback and Take-Over Requests for Level 3 Automated Vehicles

We are currently experiencing a paradigm shift towards fully automated vehicles (AVs). On the way towards fully AVs, we will experience an increase in numbers of automated vehicles on our roads, requiring the human driver to take back control in situations, which cannot be handled by the vehicle. These human-robot take-over requests (TORs) can lead to safety risks, in particular in scenarios when the driver fails to understand the TOR and, hence, lacks situational awareness (SA). In this paper, the correlation between reaction time, multi-modal feedback, informing the human driver of a transition in automation level, and success of transfer of control has been investigated. Nineteen human drivers have participated in experiments in a full-sized driving simulator: First, the driver was engaged in a secondary reading task while the car was in self-driving mode. Then, a TOR indicated to the driver to take back control. Seven different feedback modalities for the TORs have been created consisting of an audio chime, a visual cue or a static mechano-tactile haptic feedback, or a combination of these. The mechano-tactile feedback is hereby given through soft pneumatic actuators embedded into a novel soft robotic driver's seat. After the driver experienced the TOR, they were given seven seconds to regain SA, retake the driving task and react to a road incident ahead. Based on the results, it can be concluded that reaction times below 2.6 seconds and above 6 seconds result in an unsuccessful transfer of control. Additionally, we have found that haptic feedback results in a timely and safe transfer of control within a shorter time frame, when added to currently commercially available auditory and visual feedback.