Japanese scientists have found that sticking giant googly eyes in front of self-driving cars can improve pedestrian safety.
Safety is the most crucial hurdle self-driving cars have to clear before they can hit the streets. Although companies like Tesla are capturing billions of miles of real-life road data to improve its Autopilot self-driving software, technology in general still has a long way to go.
Tesla still does not classify Autopilot-equipped vehicles as self-driving cars because they still require active monitoring by the driver, who must keep their hands on the wheel and be ready to intervene at all times, according to the website. of the company.
While other cars are a big part of the challenge facing self-driving cars, pedestrians are another.
“There are not enough studies on the interaction between self-driving cars and the people around them, such as pedestrians,” said Professor Takeo Igarashi, a computer science researcher at the Graduate School of Information Science and Technology. from the University of Tokyo. Tuesday press release.
“So we need more investigation and effort on such interaction to bring security and assurance to society regarding self-driving cars.”
One way to improve safety could be to give ‘eyes’ to self-driving cars, according to a new study by Igarashi and colleagues.
This is something that various car manufacturers have explored, the idea being that the eyes would serve as a means of communication between the car and pedestrians.
Specifically, the researchers studied a situation in which a pedestrian is in a hurry to cross a street. If the car’s eyes were not looking at the pedestrian, it implied that the car was not paying attention to them and would not stop. If the eyes were looking at the pedestrian, it implied that the car recognized he was there and would stop.
If the pedestrian judged that the car could see him and would stop, he could continue to cross the street. Otherwise, the pedestrian would stop for his safety.
The researchers built such a “car” from a cart, similar in size and shape to a golf cart, with eyes. They then shot 360-degree videos of the car and imported them into a virtual reality environment.
In this VR environment, 18 participants were asked to make street crossing decisions within three seconds while the researchers studied their error rates. An error was counted if the pedestrian and car stopped, or if the car and pedestrian continued.
The results showed that the total error rate was 50.56% in the car without eyes and 29.44% in the car with eyes, which “indicates that a car with eyes can help pedestrians to make street crossing decisions more correctly than a car without eyes in general,” the researchers wrote.
While some of the participants said the eyes looked cute, others said they were scared of them despite the suggested safety upgrades.
“In the future, it would be better for a professional product designer to come up with the best design, but it would probably still be difficult to satisfy everyone,” Igarashi said in a university press release.
It’s important to keep in mind that the study was limited by a small number of participants involved in a single scenario, while the VR environment might have affected the results. Igarashi said he hopes the research will encourage similar investigations into anything that might improve interactions between self-driving cars and people.
The study, ‘Can Eyes on a Car Reduce Traffic Accidents?’, was published in AutomotiveUI ’22: Proceedings of the 14th International Conference on Automotive User Interfaces and Interactive Vehicle Applications.