Everyone is building drones, from search giant Google, to social network Facebook and retail giant Amazon. They’re all testing drones that can improve their businesses. While Facebook & Google are building drones to improve internet coverage, especially in remote areas where internet availability is very low, Amazon wants to build a drone that can deliver goods to customers.
A ”Researcher at the Massachusetts Institute of Technology’s Computer Science and Artificial Intelligence Lab have developed a drone that can ‘self-fly’ at speed of up to 30 miles per hour.This drone is a step ahead of other drones.
Back in June, the U.S. FAA-(Federal Aviation Administration) said they will finalize rules regarding the use of drones & safety of people by end of June next year. Regulators wants to avoid drone accidents that could harm a person, or a private property.
So Now, Andrew Barry has developed a software that helps drones to avoid obstacles & fly safely.This "self-fly" feature is dependent on the software.
It’s like a Google’s driverless car. This Self-Fly drone builds map of surroundings in real-time and avoid obstacles through a computer algorithm. Barry claims the new self-flying drone’s software is 20 times faster now, and its mapping tech can extract information at a speed of 8.3 milliseconds per frame, at 120 frames per second.
This project is Open-Source and its code are available online at Github.
You can watch the video below demonstrating how the drone avoid obstacles.
The main aim of this drones is to fly without hitting a wall trees and any other obstacles, while maintaining the pace.
Here’s how his drone works. It's drone algorithms that concentrates on ‘self-avoiding’ obstacles that would use pictures captured by cameras and search through the depth-field at multiple distances like one or two meters. This algorithm determines if an object is in drone’s way.
According to the MIT researcher, these algorithms are computationally intensive, and limits the speed of the drone without specialized processing hardware. He realized that the world simply doesn’t change much between frames, so his software is computing just a small subset of measurements just 10 meters away.
Barry says, “As you fly, you push that 10-meter horizon forward, and, as long as your first 10 meters are clear, you can build a full map of the world around you.”
While the project is still under development and many more features will come in near future.
The algorithm was tested using a drone with off-the-rack parts costing about $1,700.
