Autonomous vehicle system performance is limited by uncertainties inherent in the driving environment and challenges in processing sensor data. Engineers thus face the design decision of biasing systems toward lower sensitivity to potential threats (more misses) or higher sensitivity (more false alarms). We explored this problem for Automatic Emergency Braking systems in Level 3 autonomous vehicles, where the driver is required to monitor the system for failures. Participants (N=48) drove through a simulated suburban environment and experienced detection misses, perfect performance, or false alarms. We found that driver vigilance was greater for less-sensitive braking systems, resulting in improved performance during a potentially fatal failure. In addition, regardless of system bias, greater levels of autonomy resulted in significantly worse driver performance. Our results demonstrate that accounting for the effects of system bias on driver vigilance and performance will be critical design considerations as vehicle autonomy levels increase.

Autonomous vehicle systems and their users need to collaborate to navigate the driving environment, particularly during an unstructured transition of control from automation, when the system releases control and expects the human to immediately assume driving responsibility. We investigated how such transitions affect the user's trust in the system and subsequent performance. In a full-vehicle driving simulator, participants encountered two system failures: the first varied in severity (mild or severe failure), and the second required a transition of control that was either detected and alerted (loud failure) or not (silent failure). We observed (i) significant changes in user trust in the system over time and between events, and (ii) that the first failure's severity level did not affect user performance in the subsequent failure; rather, the system's detection and alert both times was sufficient to successfully complete the transition of control.

Automated emergency braking (AEB) systems-which alert a driver to approaching hazards and automatically brake-are currently available in some vehicles and will soon be widespread. Due to the uncertainties inherent in any environment and difficulties in processing sensor data, these systems are prone to both false alarms and system misses of hazards. A pressing design concern is whether to bias these systems toward a higher likelihood of false alarms versus system misses for non-fatal events. We investigated how drivers form mental models of the AEB system and how that influences their reliance on the system during a critical, potentially fatal, failure. In a full vehicle driving simulator, participants experienced nine interactions that reflected the system's level of bias toward false alarms or toward misses for non-fatal events or that demonstrated perfect performance. When a potentially fatal event occurred, participants trained to expect misses were better able to avoid a pedestrian in the road after a detection failure than those using a system with perfect performance or false alarms. These findings suggest that systems biased toward misses in non-fatal events encourage driver vigilance and preparedness for potentially fatal events.

The advent of automated features in modern vehicles requires human factors researchers to find measures other than driving behavior to anticipate the response of drivers in various contexts. Functional near-infrared spectroscopy (fNIRS) is one research tool that allows us to quantify the driver's mental state. However, the underlying mechanisms of fNIRS technology can limit the possible contexts for its application. The pervasive question arises, whether the measurement device at hand is suitable for the research topic in question and is it capable of detecting the phenomenon under investigation? We provide a proof of concept study demonstrating that significant habituation is present when drivers operate new automated driving systems and that fNIRS technology is suitable to detect said driver habituation effects. The study presented here was conducted in a driving simulator and investigated the drivers' cortical activation in three different modes of automation: manual, partially autonomous, and fully autonomous modes.

In partially automated driving, rapid transitions of control present a severe hazard. How long does it take a driver to take back control of the vehicle when engaged with other non-driving tasks? In this driving simulator study, we examined the performance of participants (N=30) after an abrupt loss of automated vehicle control. We tested three transition time conditions, with an unstructured transition of control occurring 2s, 5s, or 8s before entering a curve. As participants were occupied with an active secondary task (playing a game on a tablet) while the automated driving mode was enabled, they needed to disengage from the task and regain control of the car when the transition occurred. Few drivers in the 2 second condition were able to safely negotiate the road hazard situation, while the majority of drivers in the 5 or 8 second conditions were able to navigate the hazard situation safely.

Until vehicles are fully autonomous, safety, legal and ethical obligations require that drivers remain aware of the driving situation. Key decisions about whether a driver can take over when the vehicle is confused, or its capabilities are degraded, depend on understanding whether he or she is responsive and aware of external conditions. The leading techniques for measuring situation awareness in simulated environments are ill-suited to autonomous driving scenarios, and particularly to on-road testing. We have developed a technique, named Daze, to measure situation awareness through real-time, in-situ event alerts. The technique is ecologically valid: it resembles applications people use in actual driving. It is also flexible: it can be used in both simulator and on-road research settings. We performed simulator-based and on-road test deployments to (a) check that Daze could characterize drivers' awareness of their immediate environment and (b) understand practical aspects of the technique's use. Our contributions include the Daze technique, examples of collected data, and ways to analyze such data.

In this paper, we introduce two different transforming steering wheel systems that can be utilized to augment user experience for future partially autonomous and fully autonomous vehicles. The first one is a robotic steering wheel that can mechanically transform by using its actuators to move the various components into different positions. The second system is a LED steering wheel that can visually transform by using LEDs embedded along the rim of wheel to change colors. Both steering wheel systems contain onboard microcontrollers developed to interface with our driving simulator. The main function of these two systems is to provide emergency warnings to drivers in a variety of safety critical scenarios, although the design space that we propose for these steering wheel systems also includes the use as interactive user interfaces. To evaluate the effectiveness of the emergency alerts, we conducted a driving simulator study examining the performance of participants (N=56) after an abrupt loss of autonomous vehicle control. Drivers who experienced the robotic steering wheel performed significantly better than those who experienced the LED steering wheel. The results of this study suggest that alerts utilizing mechanical movement are more effective than purely visual warnings.

Automated driving systems that share control with human drivers by using haptic feedback through the steering wheel have been shown to have advantages over fully automated systems and manual driving. Here, we describe an experiment to elicit tacit expectations of behavior from such a system. A gaming steering wheel electronically coupled to the steering wheel in a full-car driving simulator allows two participants to share control of the vehicle. One participant was asked to use the gaming wheel to act as the automated driving agent while another participant acted as the car driver. The course provided different information and visuals to the driving agent and the driver to simulate possible automation failures and conflict situations between automation and the driver. The driving agent was also given prompts that specified a communicative goal at various points along the course. Both participants were interviewed before and after the drive, and vehicle data and drive video were collected. Our results suggest that drivers were able to interpret simple trajectory intentions, such as a lane change, conveyed by the driving agent. However, the driving agent was not able to effectively communicate more nuanced, higher level ideas such as availability, primarily due to the steering wheel being the control mechanism. Torque on the steering wheel without warning was seen most often as a failure of automation. Gentle and steady steering movements were viewed more favorably.

This tutorial is a hands-on introduction to human-centered design topics and practices for human-robot interaction. It is intended for researchers with a variety of backgrounds, particularly those with little or no prior experience in design. In the morning, participants will learn about user needs and needfinding, as ways to understand the stakeholders in research outcomes, guide the selection of participants, and as possible measures of success. We then focus on design sketching, including ways to represent objects, people and their interactions through storyboards. Design sketching is not intended to be art, rather a way to develop and build upon ideas with oneself, and quickly communicate with colleagues. In the afternoon, participants will use the tools and materials, and learn techniques for lightweight physical prototyping and improvisation. Participants will build a small paper robot (not actuated) of their own design, to practice puppeteering, explore bodily movement and prototype interactions.

While automated driving systems will become increasingly capable and common in the future, there will still be instances when human drivers want or need to make corrections to the car's automated driving behavior. We conducted two studies exploring how driving interfaces could be designed to better execute the drivers' intentions. In our first study, adult participants (N=40) experienced a simulated driving scenario that varied the behavior of the car's automation (perfect driving and imperfect driving) and the intervention modalities (takeover and takeover+influence). At certain segments, the car's automation would drive perfectly or weave within the lane. During those times, participants could intervene using the available modalities. When experiencing instances of imperfect driving, drivers who had the ability to takeover+influence intervened more often than drivers who were only given the option to takeover. As intervening would require them to resume full control, drivers in the takeover condition were more tolerant of the imperfect driving. Also, most drivers tried to intervene initially by influencing the car, even those drivers who were only given the ability to takeover. In our second study, we examined how participants (N=40) of different demographics (high school students and seniors) would respond when they were subjected to the imperfect driving scenarios. High school drivers intervened just as much as the adult drivers. However, senior drivers intervened far less. These two studies suggest that when intervention is necessary, human drivers have a desire for shared control, which allows them to act as supervisors rather than operators of automated vehicles.

With automated driving systems, drivers may still be expected to resume full control of the vehicle. While structured transitions where drivers are given warning are desirable, it is critical to benchmark how drivers perform when transition of control is unstructured and occurs without advanced warning. In this study, we observed how participants (N=27) in a driving simulator performed after they were subjected to an emergency loss of automation. We tested three transition time conditions, with an unstructured transition of vehicle control occurring 2 seconds, 5 seconds, or 8 seconds before the participants encountered a road hazard that required the drivers' intervention. Few drivers in the 2 second condition were able to safely negotiate the road hazard situation, while the majority of drivers in 5 or 8 second conditions were able to navigate the hazard safely. Similarly, drivers in 2 second condition rated the vehicle to be less likeable than drivers in 5 and 8 second conditions. From the study results, we are able to narrow in on a minimum amount of time in which drivers can take over the control of vehicle safely and comfortably from the automated system in the advent of an impending road hazard.