Flying drones is the new hit hobby, and several companies have risen to focus on this market. However, no one has yet to come up with a way to ensure that users do not constantly ram their drones into a wall or a light post.
This is about to change now, as a researcher at MIT's Computer Science and Artificial Intelligence Lab (CSAIL), managed to come up with a system that could make for fewer drone incidents in the future. It is an obstacle-detection system that allows the drone to dip, dart and overall just dodge obstacles that are in the path of the drone.
CSAIL Ph.D. student Andrew Barry, who developed the system as part of his thesis with MIT professor Russ Tedrake, explains that building drones has become ubiquitous, but no one managed to prevent drones from bumping into things and avoid obstacles.
"Sensors like lidar are too heavy to put on small aircraft, and creating maps of the environment in advance isn't practical," Barry explains. "If we want drones that can fly quickly and navigate in the real world, we need better, faster algorithms."
Any drone with the obstacle-detection system installed can avoid trees and other objects at 30-miles per hour, according to Barry. The system takes 8.3 milliseconds to process each frame, which makes it 20 times faster than any other similar system available right now.
Barry's software is also open-source and available on GitHub, so interested persons can give it a spin.
What's also interesting about this drone is that it is self-flying as opposed to human flight, something that might turn off some drone enthusiast.
For now, the software is not yet perfect; however, Barry is hoping to make changes in the future to allow for more accurate and faster obstacle detection methods. Now, since the drones are currently unmanned, we're not certain if the software can be implemented into manned drones.
Barry did not make any statement about such a plan, but it could be a good idea.
