Autonomous driving: Robot eyes can help avoid accidents with pedestrians

A study by a Japanese research team from the University of Tokyo has shown that moving robotic eyes on autonomous vehicles can increase pedestrian safety. The oversized eyes at the front of such a vehicle are movable and may or may not look at the pedestrian. This helps pedestrians make safe decisions about their next action on the road. However, there are gender-specific differences.

The decision as to whether a vehicle will stop for a pedestrian or not can ultimately be based on eye contact between driver and pedestrian. If in doubt, the pedestrian does not cross the street. With autonomous vehicles, the driver, who is more of a controller, does not have his full attention to the traffic situation. In some autonomous vehicles, there can be no controllers at all. In the first case, eye contact may not be sufficient, in the second case it may even be missing altogether, as the scientific team describes the problem examined in their study “Can Eyes on a Car Reduce Traffic Accidents?”published in Proceedings of the 14th International Conference on Automotive User Interfaces and Interactive Vehicular Applications.

So that a pedestrian can tell if a self-driving car has recognized them, the scientists equipped a self-driving golf cart with two remote-controlled large eyes similar to those in the cars in the Pixar film “Cars”. A test setup should show how risky pedestrians who are in a hurry behave when they want to cross a street just before an autonomous vehicle. The subjects were asked to decide for themselves whether the vehicle would stop or not. In the test setup, the eyes were controlled in such a way that they were sometimes directed towards the pedestrian (vehicle stops) and sometimes not (vehicle does not stop). In addition, it was tested how pedestrians behave in the same situation with a vehicle without eyes (stops or does not stop).



The photo shows the different situations in which the robot eyes were on the study participants (a) or not (b). The other two photos (c) and (d) show the situation without eyes.

To ensure that misunderstandings do not lead to an accident, the situations were filmed with 360-degree cameras and played back to the 18 study participants, who consisted of nine women and nine men between the ages of 18 and 49, using virtual reality glasses. The scenarios were played to them several times, each time in random order. They then had three seconds to decide whether or not to cross the street. From this, the scientific team determined how often the subjects made right or wrong decisions.

There were differences between the sexes: men made more dangerous decisions, meaning they wanted to quickly cross the street even though the vehicle wasn’t stopping. However, the error rate was reduced when the robot eyes did not look at them and the test subjects received the signal that the car would overlook them. In the situations where the car stopped anyway, the eyes that were fixed on the study participants didn’t make much of a difference.

The study participants, on the other hand, made “more inefficient decisions,” explains Chia-Ming Chan, one of the co-initiators of the study. For example, women behaved more cautiously and did not cross the street, even if the vehicle eventually stopped. Here, however, the gaze of the robot eyes could have a positive influence on behavior and reduce the error rate. Overall, it turned out that the view of the robot eyes made crossing the street safer for everyone.

The subjects perceived the eyes differently: some thought they were cute, others thought they were rather creepy. Women felt safer when the eyes were on them, men perceived the situation as much more dangerous when the eyes weren’t on them.

The researchers believe that the study confirms their assumptions that robotic eyes on a vehicle show pedestrians that they are seen when they look at them, and that they are overlooked when they look away. However, the science team limits the generality of the results. The number of subjects was limited. They could also have made a different decision in real life than in virtual reality.

The team now plans to develop automatically controlled robotic eyes. They could then also be different in terms of visual design in order to generate more acceptance. In any case, traffic safety in connection with autonomous vehicles is increased by the eyes, say the scientists,


(olb)

To home page

Leave a Comment