New AI model should make autonomous vehicles more accurate in bad weather

Researchers have developed an AI model that allows autonomous vehicles to more accurately determine their location on the track in changing weather conditions. Combined data from sensors that are not affected by bad weather is used for this.

According to the researchers from the University of Oxford and the Turkish University of Bogazici, bad weather such as rain, fog and snow can prevent autonomous vehicles from correctly assessing their position on the track and in relation to other road obstacles. To solve this problem, they developed an algorithm that should help determine the location of the autonomous vehicle using data from visual sensors such as cameras, and data from sensors that are less affected by bad weather conditions, such as radars.

They then trained the AI ​​model using publicly accessible datasets from autonomous vehicles and concluded after testing that the location of autonomous vehicles was better determined in bad weather. The researchers believe that this model will enable large-scale testing of autonomous vehicles. According to them, these vehicles are currently mainly being tested on a small scale.

Full Self-Driving Computer in Tesla vehicles for illustration