Assessing and Modeling Effects of Takeover Request Design on Driver Situation Awareness and Takeover Response During Scheduled Takeovers in Automated Vehicles
Restricted (Penn State Only)
- Author:
- Tan, Xiaomei
- Graduate Program:
- Industrial Engineering
- Degree:
- Doctor of Philosophy
- Document Type:
- Dissertation
- Date of Defense:
- February 21, 2023
- Committee Members:
- Steven Landry, Program Head/Chair
Ling Rothrock, Major Field Member
Yiqi Zhang, Chair & Dissertation Advisor
Myounghoon Jeon, Special Member
Sean Brennan, Outside Unit & Field Member
Scarlett Miller, Major Field Member - Keywords:
- Conditional driving automation
takeover request
scheduled takeovers
situation awareness
mathematical modeling - Abstract:
- Conditionally automated vehicles (AVs), known as the third level of driving automation, enable drivers to divert their attention from the road under specific conditions defined by the Operational Design Domain (ODD), which is a description of specific conditions (e.g., roadway features, speed range, and weather) that an AV is designed to operate adequately (NHTSA, 2017). However, when a Level 3 AV is about to exit the ODD, the driver must intervene and resume control upon a takeover request (ToR) initiated by system. The Highway Pilot feature, for instance, enables a car to drive itself in designated areas, but it will require the driver to take over and drive on roads that the Highway Pilot is not compatible with. In most cases, the takeover for ODD exit, such as exiting a freeway, can be scheduled ahead of time, rendering drivers ample time to perform the takeover task at their own pace. Despite the relatively low safety risk of scheduled takeovers, drivers still stand a chance of regaining control without acquiring necessary situation awareness (SA), which can lead to maneuver failure. Therefore, it is crucial to evaluate ToR design for ODD exit and understand its impact on driver SA restoration and takeover performance in conditionally AVs. Unfortunately, few studies have examined the design of ToR for scheduled takeover events. This dissertation aims to address this gap by focusing on scheduled freeway exiting takeover scenarios and investigating ToR designs that enhance SA restoration and takeover decision making and performance in conditionally AVs. Two experiments were conducted to obtain empirical evidence on the effectiveness of ToR designs. A web-based experiment investigated the association of ToR lead time with driver SA, takeover response time, and system evaluations to determine the most appropriate range of ToR lead time for freeway exiting takeovers in conditionally AVs. Results revealed a positive effect of a longer ToR lead time on driver SA (saturated at 16–30 s), takeover response time (no saturation), takeover readiness (no saturation), trust (saturated at 16–18 s), and acceptance (saturated at 16–20 s). However, drivers’ awareness of the remaining travel distance significantly decreased when the ToR lead time exceeded 30 s. Therefore, the study recommends a lead time of 16–30 s for better SA, which could be narrowed down to 25–30 s if considering subjective evaluations. In a driving simulator experiment, eight takeover strategies were identified for scheduled freeway exiting takeovers in conditionally AVs. These strategies were characterized by delayed gaze redirection to the road (54.2% of trials) and attention alternating between the non-driving related task (NDRT) and the road (54.2% of trials). The study tested four ToR designs with a lead time of 60 s for scheduled freeway exiting takeovers. Results showed that an initial nonverbal auditory ToR was effective in prompting drivers to redirect their attention to the road. Sending verbal auditory warnings that communicated takeover urgency at multiple stages can advance drivers’ gaze redirection for preparing takeover. Additionally, displaying a colored symbol on the car dashboard indicating the takeover urgency was helpful in reducing the time for preparing and responding to takeovers without compromising SA or vehicle control performance. Moreover, the multi-stage ToR designs also improved system evaluations of takeover readiness, trust, and usefulness while decreasing cognitive workload. Mathematical models of driver takeover response were proposed to provide a more cost-efficient approach for evaluating ToR designs and better understand the underlying cognitive mechanism of processing and responding to ToRs in scheduled takeovers. With the empirical evidence, mathematical models were developed based on the Queuing Network-Model Human Processor (QN-MHP) architecture to predict drivers’ gaze redirection time and takeover response time under different ToR designs for exiting freeways in conditionally AVs. The models were validated using empirical data and demonstrated a moderate-to-good ability to account for 70-97% of the data. In summary, this dissertation presents a pioneering examination of the ToR design for scheduled takeover events in conditionally AVs. The empirical findings provide practitioners guidance on the lead time and multi-stage warning design of scheduled ToRs for freeway exiting takeovers. Moreover, the proposed mathematical models provide a cost-efficient approach for predicting driver takeover response in scheduled takeover scenarios, providing a foundation for future investigations into other scheduled ODD exit takeover scenarios and the incorporation of other ToR characteristics into the models.