NVIDIA has taught the drone to navigate in space without the use of GPS
As you know, GPS systems are among the most common and accurate source of coordinates. They are used everywhere, including in unmanned aerial vehicles. But still there are areas where satellite navigation systems are not working, such as woods, tunnels, caves and so on. The researchers therefore NVIDIA decided to develop its own version of the system for navigation of robots, which would not depend on GPS.
The development of world famous company relies on the operation of the camera under control of artificial intelligence for recognition of incoming data and decision-making. During testing one of the production models of the drone was equipped with two additional cameras and a module Jetson TX1, which bore the brunt of the function artificial intelligence and deep machine learning.
The new system is designed for unmanned aerial vehicles were able to navigate difficult given the space to plot a route and navigate without assistance. It was originally planned that the drones with such a system can be used only in the forest, but, as shown, they are suitable for use in the city, in areas of natural disasters and even (with slight modifications) under water.
The neural network which controls the action of the aircraft, called TrailNet, and her training took place not only in the field, but using the footage from the GoPro taken in forest environment, various types, starting from forests in North America to forests on the slopes of the Alps. The total distance of the journey, captured on camera, made almost 13 000 km. Now the drone may already own through the forest at a distance of 1 kilometer thanks to the advanced functions of anti-collision. As stated by one of the developers of the new system Nicholas Smolyan,
“The visual system will work where no GPS or other navigation system. Everything needs this system is only in the presence of any recognisable landmarks”.
