Objective
This study develops a computational model to predict drivers’ response time and understand the underlying cognitive mechanism for freeway exiting takeovers in conditionally automated vehicles (AVs).
Background
Previous research has modeled drivers’ takeover response time in emergency scenarios that demand a quick response. However, existing models may not be applicable for scheduled, non-time-critical takeovers as drivers take longer to resume control when there is no time pressure. A model of driver response time in non-time-critical takeovers is lacking.
Method
A computational cognitive model of driver takeover response time is developed based on Queuing Network-Model Human Processor (QN-MHP) architecture. The model quantifies gaze redirection in response to takeover request (ToR), task prioritization, driver situation awareness, and driver trust to address the complexities of drivers' takeover strategies when sufficient time budget exists.
Results
Experimental data of a preliminary driving simulator study were used to validate the model. The model accounted for 97% of the experimental takeover response time for freeway exiting.
Conclusion
The current model can successfully predict drivers’ response time for scheduled, non-time-critical freeway exiting takeovers in conditionally AVs.
Application
This model can be applied to the human-machine interface design with respect to ToR lead time for enhancing safe freeway exiting takeovers in conditionally AVs. It also provides a foundation for future modeling work towards an integrated driver model of freeway exiting takeover performance.
more »
« less
This content will become publicly available on October 19, 2024
Predicting Driver Takeover Decisions in Conditionally Automated Vehicles with a Gaze-based Deep Learning Model
This study proposes a novel methodology for modeling driver takeover behavior in conditionally automated vehicles (AVs) when exiting a freeway using deep learning (DL) network architectures. While previous research has focused on modeling takeover time in emergency scenarios, which require quick responses, these models may not be applicable to scheduled, non-time-critical takeovers. In such situations, drivers may employ varying strategies and take longer to resume control of the vehicle when there is no time pressure. To address this problem, a deep learning architecture based on a convolutional neural network (CNN) was implemented to predict drivers’ takeover behaviors in scheduled takeovers. The model was trained on drivers’ driving data and eye gaze with varying time windows, facilitating an analysis of drivers’ takeover decisions to various takeover request designs. The model achieved good performance metrics, with an F1 Score of 0.993, a recall of 0.996, and a precision of 0.991. The application of these models holds substantial potential for refining the design of the human-machine interface, specifically in calibrating the takeover request (ToR) lead time, thereby promoting safe freeway exiting takeovers in conditionally AVs.
more » « less- NSF-PAR ID:
- 10470042
- Publisher / Repository:
- SAGE Publications
- Date Published:
- Journal Name:
- Proceedings of the Human Factors and Ergonomics Society Annual Meeting
- Volume:
- 67
- Issue:
- 1
- ISSN:
- 1071-1813
- Format(s):
- Medium: X Size: p. 657-663
- Size(s):
- ["p. 657-663"]
- Sponsoring Org:
- National Science Foundation
More Like this
-
-
Conditionally automated vehicles require the out-of-the-loop driver to intervene when the system is unable to handle forthcoming situations, such as freeway exiting. The takeover request (ToR) for exiting a freeway can be scheduled in advance. Upon a ToR, the driver needs to gain situation awareness (SA) and resume manual control. This study examined how the ToR lead time affects driver SA for resuming control and when to send the ToR is most appropriate for freeway exiting. A web-based, supervised experiment was conducted with 31 participants. Each participant experienced 12 levels of ToR lead time (6, 8, 10, 12, 14, 16, 18, 20, 25, 30, 45, and 60 s). The results showed positive effects of longer ToR lead times (16–60 s) on driver SA for resuming control to exit from freeways in comparison to shorter ToR lead times (6–14 s), and the effects level off at 16–30 s.more » « less
-
In highly and fully automated vehicles (AV), drivers could divert their attention to non-driving-related activities. Drivers may also take over AVs if they do not trust the way AVs drive in specific driving scenarios. Existing models have been developed to predict drivers’ takeover performance in responding to takeover requests initiated by AVs in semi-AVs. However, few models predicted driver-initiated takeover behavior in highly and fully AVs. The present study develops an attention-based multiple-input Convolutional Neural Network (CNN) to predict drivers’ takeover intention in fully AVs. The results indicated that the developed model successfully predicted takeover intentions of drivers with a precision of 0.982 and an F1-Score of.989, which were found to be substantially higher than other machine learning algorithms. The developed CNN model could be applied in improving the driving algorithms of the AV by considering drivers’ driving styles to reduce drivers’ unnecessary takeover behaviors.more » « less
-
Trust calibration poses a significant challenge in the interaction between drivers and automated vehicles (AVs) in the context of human-automation collaboration. To effectively calibrate trust, it becomes crucial to accurately measure drivers’ trust levels in real time, allowing for timely interventions or adjustments in the automated driving. One viable approach involves employing machine learning models and physiological measures to model the dynamic changes in trust. This study introduces a technique that leverages machine learning models to predict drivers’ real-time dynamic trust in conditional AVs using physiological measurements. We conducted the study in a driving simulator where participants were requested to take over control from automated driving in three conditions that included a control condition, a false alarm condition, and a miss condition. Each condition had eight takeover requests (TORs) in different scenarios. Drivers’ physiological measures were recorded during the experiment, including galvanic skin response (GSR), heart rate (HR) indices, and eye-tracking metrics. Using five machine learning models, we found that eXtreme Gradient Boosting (XGBoost) performed the best and was able to predict drivers’ trust in real time with an f1-score of 89.1% compared to a baseline model of K -nearest neighbor classifier of 84.5%. Our findings provide good implications on how to design an in-vehicle trust monitoring system to calibrate drivers’ trust to facilitate interaction between the driver and the AV in real time.more » « less
-
The rapid growth of autonomous vehicles is expected to improve roadway safety. However, certain levels of vehicle automation will still require drivers to ‘takeover’ during abnormal situations, which may lead to breakdowns in driver-vehicle interactions. To date, there is no agreement on how to best support drivers in accomplishing a takeover task. Therefore, the goal of this study was to investigate the effectiveness of multimodal alerts as a feasible approach. In particular, we examined the effects of uni-, bi-, and trimodal combinations of visual, auditory, and tactile cues on response times to takeover alerts. Sixteen participants were asked to detect 7 multimodal signals (i.e., visual, auditory, tactile, visual-auditory, visual-tactile, auditory-tactile, and visual-auditory-tactile) while driving under two conditions: with SAE Level 3 automation only or with SAE Level 3 automation in addition to performing a road sign detection task. Performance on the signal and road sign detection tasks, pupil size, and perceived workload were measured. Findings indicate that trimodal combinations result in the shortest response time. Also, response times were longer and perceived workload was higher when participants were engaged in a secondary task. Findings may contribute to the development of theory regarding the design of takeover request alert systems within (semi) autonomous vehicles.more » « less
