Autonomous cars equipped with giant eyes would reduce the risk of accidents with pedestrians

⇧ [VIDÉO] You might also like this partner content (after ad)

Every year, about 15% of people killed on French roads are pedestrians. The problems of reciprocal visibility and anticipation between pedestrians and drivers are largely responsible, which raises real questions in terms of safety in the age of driverless autonomous cars. Recently, a team of Japanese researchers developed a concept of a “watching” car with robotic eyes, allowing the pedestrian to understand the behavior of the car. This innovation could equip future autonomous cars, the design remains to be defined.

The development of autonomous vehicles quickly brings us to a new era of transport. Whether delivering parcels, plowing fields or transporter of people within a campus, a great deal of research is underway to concretely implement this concept, while ensuring the road safety.

But one of the main differences between ordinary and autonomous vehicles is that the drivers of the latter can become more passengers, so that they no longer pay their full attention to the road, or that there is no – simply being a nobody behind the wheel. This makes it difficult for pedestrians to assess whether a vehicle has noticed their presence or not, as there may be no eye contact or indication of people inside.

This is why researchers from the University of Tokyo have studied more precisely this concern affecting the very relationship to the autonomous car, which should be more “human”. They wanted to find a method to notify pedestrians when an autonomous vehicle has noticed them and intends to stop, or if it hasn’t seen them and therefore won’t stop when they pass. A paper on their research was presented at the 14th International Conference on Automotive User Interfaces and Interactive Vehicle Applications.

A car straight out of a cartoon

In order to overcome the problem of “connection” between the car and the pedestrian, the researchers equipped an autonomous golf cart with two large robotic eyes operated manually by the driver, like a cartoon car. Being fitted with a one-way reflective film on the windshield, the interior of the vehicle escaped the view of pedestrians, giving the impression that there was no driver. The researchers called it ” The Gazing Car (watching the car). They wanted to test whether putting wiggly eyes on the vehicle would affect people’s risky behavior, i.e. whether people would always cross the road in front of a moving vehicle when in a hurry.

Takeo Igarashi of the Graduate School of Information Science and Technology says in a communiqué : « There are not enough investigations into the interaction between self-driving cars and the people around them, such as pedestrians. We therefore need more investigation and effort on such interaction to provide security and assurance to society regarding self-driving cars. ».

The team set up four scenarios, two where the vehicle had eyes and two where it didn’t. Either the car had noticed the pedestrian and intended to stop, or it hadn’t noticed and continued to drive. When equipped with eyes, they either looked towards the pedestrian, to signify ” I will stop “, or elsewhere, meaning ” I won’t stop ».

In the experiment, the participants have to decide if the vehicle has noticed them and will stop. Images show a participant’s first-person view. In (a) the vehicle pays attention to the participant who is safe to cross; in (b) he does not pay attention to the participant, who considers it dangerous to cross; and in (c) and (d), the participant does not know. © Chang et al. 2022.

Conveniently, 18 volunteers — nine women and nine men, ages 18 to 49, all Japanese — participated in the study. The authors decided to test their attitude thanks to augmented reality, so as not to take risks in a real situation.

This is how the team recorded the scenarios using 360-degree video cameras. Thanks to augmented reality, the volunteers safely experienced the scenarios 40 times in random order and had three seconds each time to decide whether or not they would cross the road in front of the vehicle. The researchers recorded their choices and measured error rates, i.e. how often they chose to stop when they could have crossed and how often they crossed when they should have waited.

Video of the virtual reality experience in which participants cross a road on the University of Tokyo campus. © Chang et al. 2022.

Different results between men and women

After analyzing the results, a clear difference in reactions appears between the sexes. Chia-Ming Chang, lecturer and member of the research team explains: Although other factors such as age and background also influenced participants’ reactions, we believe this is an important point, as it shows that different road users may have different behaviors and needs, which require different means of communication in our future self-driving world ».

Concretely, men made the greatest number of dangerous decisions to cross the road (i.e. choosing to cross when the car did not stop), but these errors were reduced by looking of the vehicle. On the women’s side, the authors point out that they made ineffective decisions. In other words, they chose not to cross when the car intended to stop. But these errors were also reduced by the look of the vehicle.

However, the presence of the eyes did not significantly change the reactions of men in safe situations (i.e. choosing to cross when the car was about to stop) or of women in dangerous situations. . Indeed, they did not cross whether the vehicle had eyes or not. The authors conclude that the eyes result in a smoother or safer crossing for everyone.

A reaction induced by a humanized car

The authors then wondered what the eyes made the participants feel, to explain the different results obtained. The feelings are just as mixed as the previously mentioned reactions.

Thus, some volunteers found them (the eyes) cute, while others saw them as scary. For many male participants, when their eyes were averted, they reported feeling the situation was more dangerous. For women, when the eyes looked at them, many said they felt safer.

Igarashi says, ” We focused on eye movement, but didn’t pay too much attention to their visual design in this particular study. We kept it simple to minimize design and construction costs due to budget constraints. In the future, it would be better for a professional product designer to find the best design, but it would probably still be difficult to satisfy everyone. ».

The team acknowledges that this study is limited by the small number of participants playing a single scenario. It’s also possible that people make different choices in virtual reality than in real life. However, the change of era that is looming for the future, with these autonomous cars, requires additional safety points and the implementation of such studies will make it possible to make road use safer for all users, whether whether they are motorized or not.

The researchers point out: In the future, we would like to develop autonomous AI-connected robotic eye automatic control (instead of manually controlled), which could adapt to different situations. ».

Source: AutomotiveUI ’22: Proceedings of the 14th International Conference on Automotive User Interfaces and Interactive Vehicular Applications