People go to forest rather frequently – to hunt, to collect mushrooms or berries, or simply to take a walk. Unfortunately, some of these people get lost and it is a great challenge to find them. Now international team of scientists developed drone software that allows drones autonomously detect and follow forest paths to find lost people quicker.
And time in these situations is critical because people may starve, have injuries or simply get too cold. In Switzerland alone emergency centres respond to around 1,000 calls annually to help lost and injured people in the forests, which is why innovative solutions are needed.
Drones are already in use, but if they worked autonomously they could be used in parallel with rescue teams to accelerate the search for people lost in the wild. Professor Davide Scaramuzza from the University of Zurich said: “While drones flying at high altitudes are already being used commercially, drones cannot yet fly autonomously in complex environments, such as dense forests. In these environments, any little error may result in a crash, and robots need a powerful brain in order to make sense of the complex world around them”.
The new software relies on simple camera setup. Couple of cameras on the drone detect man-made trails using complex artificial-intelligence algorithms. It is a very difficult task as people themselves sometimes struggle to find the trail.
Scientists use so-called Deep Neural Network, which allows the drone to actually learn from experience. To start with, team were walking around in forests taking pictures of trails with cameras attached to a helmet. They created a database of these images and allowed the drone to use it. When drone was tested in a completely new area, it found the correct direction in 85% of cases. For comparison, humans could only do it in 82% of the cases.
Scientists have been working on deep learning in neural networks since the early 1990s. However, a lot still needs to be done till drones can autonomously help looking for lost people. One of the tasks is to teach drones to recognize humans. The ultimate result is worth the effort – eventually robots will be working with humans hand by hand in searching and rescuing missions.