Swiss researchers have come up with a rescue drone which can autonomously identify and follow passageways via the undergrowth and rough terrain, with the help of sophisticated artificial intelligence.

The ultimate envision of the scientists is an entire fleet of the rescue drones that can swarm forests seeking missing people, functioning along with their human counterparts.

Davide Scaramuzza, who led the project for the Dalle Molle Institute for Artificial Intelligence and the University of Zurich, said, “While drones flying at high altitudes are already used commercially, drones cannot yet fly autonomously in complex environments, like dense forests. In these environments, any little error may result in crash, and robots need powerful brain to make sense of complex world around them”.

According to researchers, their latest breed of quadrocopter may get rapidly deployed in huge numbers to complement human rescue teams, cutting the response time. The device depends on tiny cameras and unique software.

Fellow researcher Alessandro Giusti explained that instead of depending on sophisticated sensors, quite powerful artificial-intelligence algorithms are used by the drone to interpret the pictures to identify man-made trails. In case a trail is evident, the software steers the drone in the corresponding direction.

They said that the Swiss team has used a ‘deep neural network’ that triggers a computer to learn to solve complicated tasks from a group of training examples, similar to a way the brain learns from experience. The study was incredibly complex as they hiked along a number of trails in the Swiss Alps, and captured more than 20,000 pictures of the environment using a helmet-mounted camera.

Mr. Giusti reported in his analysis that the effort has paid off as when they tested on a latest, previously unseen trail, the deep neural network was found the right direction in 85% of cases. In comparison to that when humans faced with the same task, they guessed rightly 82% of the time.