Andrew Barry, an MIT Phd candidate, proposes a new approach in drone construction.
In the race to create more powerful and intelligent flying drones, a student from the Massachusetts Institute of Technology took the lead in innovation. High-speed obstacle-detection drone was created at MIT as part of Andrew Barry’s PhD thesis.
Barry with the help of the Computer Science and Artificial Intelligence Lab managed to develop a software for drones making them capable of going to speeds up to 30 miles per hour without them bumping into various obstacles.
The student said that the system will soon be capable of doing loops in and through a heavy treed area without even flinching. It seems that Barry’s research endeavor was prompted by the drone’s navigational limitations.
Professor Russ Tedrake, Barry’s PhD tutor, expressed his view on drones by saying that they are indeed a marvel of modern technology and everybody is rushing into building them in large number. But there is one thing they have missed: making them not bump into stuff.
So what was Barry’s approach on creating a faster and more intelligent flying drone? He said that the answer lies in the number of equipment the drone’s fitted with. LIDAR type sensors are much too heavy to put on a drone. The answer may lie in the way the drone is able to map its surroundings. Barry said that in order to solve this problem we don’t need to feed the drone additional charts, created before his first flyby. This could only put a strain on the drone’s computational capabilities, thus slowing it down.
He also added that the only thing that can give a drone that certain edge in mid-air is a powerful and fast algorithm. Basically, a common algorithm would use information captured via the drone’s imagining sensor in order to search the field at certain distances. This drone uses the collected data and basically computes a map of its surrounding every 2 or three meters.
The high-speed obstacle-detection drone was created at MIT in order to demonstrate that you can easily make a drone more reliable by tweaking its programming rather than fitting it with heavy and expensive mapping instruments.
Barry proposes a different approach to this matter. Instead of continuously scanning its surrounding and creating a map every 2 or three meters, the drone would simply scan the surrounding everyten meters. Barry’s theory is that between the 10-meter frames the world does not change around the drone. By doing this simply task every 10 meters, the drone is capable of creating an accurate map of its surroundings without having to slow down.
Image source:www.pixabay.com
