The Latest Study Explores How AR Enhances Safety in Autonomous Vehicles
One important concern for future road safety with the increasing prevalence of self-driving cars is how to accommodate passengers’ wishes to relax while still keeping an eye out for potential dangers and being prepared to take over the wheel.
The University of Glasgow’s research team has been exploring the possibility of using augmented reality technology to let drivers experience the advantages of autonomous vehicles while still having the ability to take control of the car swiftly if needed.
Their findings provide credence to the idea that augmented reality heads-up displays (HUDs) might improve drivers’ ability to utilise entertainment applications without losing focus on the road by superimposing eye-catching images over real-world vistas seen through automobile windscreens. Careful design management is necessary to avoid information overload during critical moments.
As we approach the fully autonomous control that manufacturers have promised, their research will help shape the safety features of the next generation of self-driving automobiles.
Researchers at the University of Glasgow are exploring how augmented reality (AR) can enhance the autonomous driving experience, allowing drivers to engage in other activities while being prepared to take control when necessary. They discovered that using AR heads-up displays to project captivating graphics on car windscreens can help maintain road awareness while using entertainment apps, although careful design is essential to avoid information overload during critical moments.
These insights are shaping the development of safety features for future self-driving vehicles as they progress towards complete automation. Thomas Goodge, a co-author from the University’s School of Computing Science, explained that autonomous vehicles offer the prospect of diverting attention from driving to other activities like reading, gaming, or watching TV. However, drivers must remain alert to road conditions to intervene in emergencies, a situation that may persist until vehicles can fully operate without human input.
Presently, drivers are not fully active in controlling the vehicle but must still monitor their surroundings to ensure safe travel. Goodge noted that the challenge with human operators in such supervisory roles is their tendency to become bored or distracted, potentially leading to dangerous delays in responding to sudden road changes. The study aimed to investigate whether AR could help occupants engage in non-driving tasks while swiftly refocusing during critical moments.
In their lab experiment, participants sat in a simulated driving setup and engaged with tasks displayed on a tablet or through an AR overlay, simulating situations they might encounter while driving. The tasks included a simple game and a more complex activity involving number entry. The simulations unexpectedly paused before potential hazards to assess situational awareness, and participants predicted the outcomes. Although participants maintained some road awareness, their prediction accuracy dropped when multitasking.
The experiment also tested AR visual cues that highlighted upcoming road conditions, which improved participants’ situational awareness, particularly during simpler tasks. Another co-author, Professor Frank Pollick, underscored the importance of tackling the “look but fail to see” phenomenon, which leads to the overlooking of crucial details despite their visibility. AR technology proved effective in mitigating this issue by focusing drivers’ attention where needed.
Professor Stephen Brewster, who leads the Multimodal Interaction Group and co-authored the study, highlighted the inherent risks of distraction in autonomous vehicles. He suggested that their findings indicate a balance where occupants can stay engaged in tasks while remaining informed about evolving road conditions. He stressed the need for further research to optimise AR support for future autonomous vehicle users.
The team’s paper, ‘Can You Hazard a Guess?’ will detail their findings. The team will present Evaluating the Effects of Augmented Reality Cues on Driver Hazard Prediction at the upcoming CHI Conference on Human Factors in Computing Systems. The research received support from the UKRI Centre for Doctoral Training in Socially Intelligent Artificial Agents and the European Research Council.
In their controlled environment, the research team positioned participants behind the wheel of a simulated vehicle as they watched on computer screens what it would be like to look out the window. Using either a tablet screen—like the big dashboard displays of modern vehicles—or an augmented reality display—which superimposed tasks on top of the road scene—the screens showed a sequence of forty video clips of actual road conditions and instructed participants to do assignments.
One activity was a basic game in which players “collected” jewels as they moved around the screen. In the second round, players had to accurately copy a phone number using more advanced number pad software.
In both cases, the video paused just before a potentially dangerous event, such as a pedestrian crossing the street, to assess situational awareness. After the movie paused, the participants were to use their knowledge of the road conditions to choose one of four possible future outcomes.
A preliminary control experiment let them make predictions without actually doing anything, so we could see how well they could forecast the hazard’s result. Participants were able to keep a partial awareness of the road conditions while utilising either the tablet for the heads-down component of the assignment or the augmented reality headgear for the heads-up portion. But in both cases, they were far less able to accurately foretell what was going to happen next.
As part of an additional experiment, just a few seconds before the video paused, the augmented reality headset positioned certain visual signals in the participants’ field of vision to alert them to impending danger on the road. In this case, participants demonstrated heightened vigilance about driving conditions; nonetheless, their performance improved more in the gem-collecting game compared to the more taxing keypad app.
In their controlled environment, the research team positioned participants behind the wheel of a simulated vehicle as they watched on computer screens what it would be like to look out the window. Using either a tablet screen—like the big dashboard displays of modern vehicles—or an augmented reality display—which superimposed tasks on top of the road scene—the screens showed a sequence of forty video clips of actual road conditions and instructed participants to do assignments.
One activity was a basic game in which players “collected” jewels as they moved around the screen. In the second round, players had to accurately copy a phone number using more advanced number pad software.
In both cases, the video paused just before a potentially dangerous event, such as a pedestrian crossing the street, to assess situational awareness. After the movie paused, the participants were to use their knowledge of the road conditions to choose one of four possible future outcomes.
A preliminary control experiment let them make predictions without actually doing anything, so we could see how well they could forecast the hazard’s result. Participants were able to keep a partial awareness of the road conditions while utilising either the tablet for the heads-down component of the assignment or the augmented reality headgear for the heads-up portion. But in both cases, they were far less able to accurately foretell what was going to happen next.
As part of an additional experiment, just a few seconds before the video paused, the augmented reality headset positioned certain visual signals in the participants’ field of vision to alert them to impending danger on the road. In this case, participants demonstrated heightened vigilance about driving conditions; nonetheless, their performance improved more in the gem-collecting game compared to the more taxing keypad app.
Can You Hazard a Guess?’ was the title of the team’s paper. Next month, at the CHI conference on Human Factors in Computing Systems, the researchers will discuss their findings on “Evaluating the Effects of Augmented Reality Cues on Driver Hazard Prediction.”
The National Science Foundation (NSF) and the European Research Council (ERC) provided financial backing for the study.
