Soon, your car may be able to automatically spot that mineshaft of a pothole in the road that never seems to get fixed.
Now, artificial intelligence can allow cars to spot potholes on the road ahead thanks to researchers at the Korea Institute of Civil Engineering and Building Technology.
“Poor road-surface conditions pose a significant safety risk to vehicle operation, especially in the case of autonomous vehicles,” the researchers wrote in their study, published in Electronics. A pothole could cause the driver — or the self-driving car — to lose control of the vehicle, resulting in injury or death.
This means that road maintenance — which is already subpar in many communities — will become even more crucial in the future, especially if we ever want self-driving cars and trucks dominating the streets.
The cost of potholes: Potholes are not just a safety issue; the axel-snappers cost drivers about $3 billion annually in repairs in the US, AAA estimated in 2016.
Poor road surfaces can be especially problematic for autonomous vehicles.
Potholes are exacerbated in rainy conditions, the researchers note. Extreme weather, including heavier and more intense rains, are one of the impacts of climate change — Seoul, for example, suffered a record-setting deluge in August of 2020. So, we’re likely to see this cost increase over time.
Spotting damages: Road conditions can be tracked by a variety of methods, including vibration sensors and laser scanning. But deep learning algorithms have recently given researchers the horsepower to begin developing AI models that can help our cars spot potholes and other damage using computer vision.
Models that use images can be done with consumer devices, like smartphones or, in this case, cars (but not with smartphones while driving your car!). This makes photo-based systems appealing to municipalities already currently relying on people to spot potholes and other problems.
But the seemingly simple task of spotting a hole in the road is a more complex problem for a computer than you may think. The open road is a dynamic environment, after all, and when the weather impacts the lighting conditions on the road, the AI may perform poorly, even on a road it has learned before. Brightness levels are a particular sticking point.
To help solve the problem, the researchers used an AI model called a convolutional neural network, or CNN. The CNN tweaks the brightness of the images the sensor sees before the images hit the AI that says ‘yep, those are potholes.’
Armed with the CNN, their AI worked regardless of the lighting.
Local governments in Korea are piloting the technology, and the researchers hope to expand that number in the near future.
“It is essential to maintain road facilities in good condition in the coming era of autonomous vehicles,” lead researcher Seungki Ryu said.
“This AI-based technology will make effective road surface management much easier.”
We’d love to hear from you! If you have a comment about this article or if you have a tip for a future Freethink story, please email us at firstname.lastname@example.org.