Researchers report that self-driving cars struggle to manage social interactions on the road, such as whether to give way to something or someone on the road or to keep going.
Should I go or give in?
It’s one of the most fundamental road traffic questions that people typically ask quickly and intuitively, relying on social interactions that are trained from the start.
According to the new study, however, self-driving cars are not as powerful.
An analysis of videos uploaded by YouTube users of self-driving cars in various traffic situations shows that self-driving cars have a particularly hard time understanding when to give way.
“The ability to navigate traffic is based on much more than just traffic rules.
Social interactions, including body language, play a large role when we signal each other in traffic.
This is where self-driving car programming still comes in,” says Barry Brown, a professor at the University of Copenhagen who has been studying the evolution of self-driving car road behavior for the past five years.
“As a result, it’s difficult for them to always understand when to stop and when someone is stopping for them, which can be both annoying and dangerous.”
“Sorry self-driving car!”
Companies like Waymo and Cruise have introduced self-driving car taxi services in parts of the United States.
Tesla has rolled out its FSD (fully autonomous driving) model to around 100,000 volunteer drivers in the US and Canada. And the media is full of stories about how well self-driving cars are working.
But according to Brown and his team, their actual road performance is a closely guarded trade secret that very few have access to. Therefore, the researchers conducted in-depth analysis using 18 hours of YouTube footage filmed of enthusiasts testing cars from the back seat.
One of her video samples shows a family of four standing on the side of a residential street in the United States.
There is no crosswalk, but the family wants to cross the street. As the driverless car approaches, it slows, causing the two adults in the family to wave their hands to signal the car to move on.
Instead, the car stays right next to them for 11 seconds. Then, as the family begins to cross the street, the car starts moving again, causing her to jump back onto the pavement, at which point the person in the back seat rolls down the window and yells, “Sorry, self-driving.” Automobile!”
“The situation is similar to the main problem we found in our analysis, demonstrating the inability of self-driving cars to understand social interactions in traffic.
The self-driving vehicle stops to avoid hitting pedestrians, but still crashes into them because it doesn’t understand the signals. Not only does it create confusion and waste time in traffic, it can also be downright dangerous,” says Brown.
The social element of driving
In tech-centric San Francisco, the performance of self-driving cars can be judged up close. Here driverless cars have been used as buses and taxis in several parts of the city, navigating the hilly streets between people and other natural phenomena.
And that, according to the researcher, caused a lot of resistance among the city’s residents.
“Self-driving cars are causing congestion and problems in San Francisco because they react inappropriately to other road users.
Recently, the city’s media wrote about a chaotic traffic incident caused by self-driving cars due to fog. Fog caused the self-driving cars to overreact, stopping and blocking traffic, even though fog is extremely common in the city,” says Brown.
Robot cars have been in development for a decade and the industry behind them has spent over 40 billion DKK (about US$5.7 billion) to speed up their development.
However, the result is that cars still make a lot of mistakes, blocking other drivers and disrupting the smooth flow of traffic.
Why is it so difficult to program self-driving cars to understand social interactions on the road?
“I think part of the answer is that we take the social element for granted. We don’t think about it when we get in a car and drive – we just do it automatically,” says Brown.
“But when it comes to designing systems, you have to describe everything that we take for granted and incorporate it into the design. The automotive industry could learn from a more sociological approach.
Understanding the social interactions that are part of transportation should be used to shape self-driving car interactions with other road users, similar to how research has helped improve the usability of cell phones and technology in general.”
The researcher presented the study at the 2023 CHI Conference on Human Factors in Computing Systems.
Source: University of Copenhagen.