No matter how many cameras or sensors they use, autonomous cars will never be perfect, but research coming out of Japan suggests that a simple upgrade could help reduce the risk of self-driving cars hitting undetected pedestrians: a pair of animated googly eyes that make it obvious what the vehicle has or hasn’t spotted.
By the time you’re old enough to be out on your own, the act of making eye contact with the driver of an approaching vehicle and ensuring the car is slowing to a stop when you’re trying to cross the road is nearly a subconscious act. But with autonomous cars, which rely on cameras and sensors hidden all over the vehicle, there’s no obvious indication as to which obstacles have been detected and will be treated as a hazard requiring a stop.
In a paper recently presented at the 14th International Conference on Automotive User Interfaces, researchers from the University of Tokyo and Kyoto University detail an experiment where a golf cart with a human but non-visible driver was upgraded with an over-sized pair of manually-controlled animated eyes on the front that could be made to look around. Researchers recreated a total of four scenarios in which the vehicle approached pedestrians intending to cross a street: one where the person was noticed and the cart was intending to stop, one where they weren’t noticed and the cart was going to keep driving, and then the same two recreated with the eyes added to the cart, looking at and away from the pedestrian in each instance.
To keep the experiment safe for participants, the scenarios were filmed with a 360-degree camera, and then 18 subjects — nine women and nine men aged 18 to 49 years — took over the role of the pedestrian through a virtual reality headset where they randomly played through the scenarios multiple times and had just three seconds to assess the situation and decide if they were going to attempt to cross the road in front of the approaching golf cart.
The researchers were surprised to find that male participants tended to make more dangerous decisions about crossing the road, choosing to cross when the cart made no indication it was going to stop, while female participants erred on the side of caution, often choosing not to cross when the cart was actually coming to a stop. However, in both instances, the participants noted that when the vehicle had eyes that were looking away, crossing felt less safe, but when the eyes appeared to be paying attention to them, crossing felt safer.
The participants also noted that the giant eyes were kind of creepy, and even a bit scary, and the researchers, who primarily focused on the eyes’ movements for this experiment, believe their effectiveness could have been increased even further if they had been designed with aesthetics in mind too, including a more lifelike appearance and movements.
The experiment is a reminder that as we transition to the next generation of transportation, changes won’t be limited to the vehicles themselves. We already equip nearly silent electric cars with added noises and warning sounds to make them more noticeable to pedestrians who are used to listening for the engine sounds of an approaching vehicle. And as weird as adding moving eyes to the front of a vehicle may seem, humans are one of the most unpredictable things a self-driving car has to deal with, and a couple of oversized googly eyes could be a relatively cheap way to help improve safety.