Self-driving cars have trouble seeing at night or in fog – but mimicking the human brain can make them safer

Imagine this: You’re driving on a mountain road when you suddenly come across a thick patch of fog. You react instinctively. Your vision sharpens and you narrow your eyes to make out the shapes of oncoming cars.

Humans cope very well with these rapid changes, but if it were a self-driving car – at least one with a current artificial intelligence (AI) system behind the wheel – everything could easily end in disaster.

Today’s AI image processing systems are extremely accurate when visibility is good. On a clear, sunny day, a self-driving car can accurately detect pedestrians, traffic signs and other vehicles. However, they are extremely vulnerable to environmental changes. When it rains, gets dark or foggy, standard AI systems become blind and can no longer detect obstacles that a human driver would easily see.

Our research at the University of Valencia suggests a possible solution: Instead of exposing AI models to millions of images of all possible road conditions,…