Drones have entered a new level of advancement. The Massachusetts Institute of Technology’s Computer Science and Artificial Intelligence Laboratory (CSAIL) has successfully developed a system that lets drones detect obstacles in their flight path and therefore avoid them.

In test flights, this has allowed the drone to fly autonomously through trees and an open field at speeds of more than 30 miles an hour (48 kph). And this was achieved using only a pair of cameras and two processors no more powerful than the ones being used in a smartphone. Check out the video below to see it in action.

Improving drones’ functionalities

They developed the obstacle scanning system with one goal in mind: figure out how to make drones that do not collide with obstacles in their flight path. Currently the public fascination for creating drones is limited to the construction itself; the improvement of drones’ functionalities isn’t something that many are paying that much attention to yet. This where they hope their system can fill the gap.

The scanner system guides the drone in detecting obstacles in its flight path by using a stereo-vision algorithm that projects a full map of the surroundings in real-time. The algorithm operates at 120 frames per second, which allows it to compute small subsets of measurements to create the map. In the case of the drone, the algorithm is set to measure distances of 10 meters away.

Traditional scanner systems operate in the same way. Using a camera, the system will calculate and search through the depth-field at multiple distances to determine if there is an obstacle in the way. This method however is computationally intensive and only allows the drone to fly at speeds of 5 to 6 miles an hour.

Intelligent flight

With their scanner system that runs 20 times faster than traditional scanners, this limitation is eased out of the equation. That is why the drone that has been equipped with their scanner system can fly faster through an obstacle-filled course in a seamless manner. The scanner system makes for a dynamic and reactive machine that can hypothetically fly through even the most challenging of environments may it be forests or the urban jungle.

For now the scanner system is limited by the computing power of its processors. All told however, this is an exciting development that presents plenty of practical uses in the future. From delivery to security, autonomous drones capable of intelligent flight may soon be a common sight in the skies.

4 Comments
  1. Drones are a topic I am very interested in. The idea of drones becoming autonomous is thrilling as you can have the drone fly its route, charge itself, and then fly its route again, without human intervention. Can’t wait to see what’s next.

  2. When this is perfected online retailers who want to use drones to deliver packages to their customers will rest assured that their buyers will get what they paid for.

    Ehang, the Chinese company building autonomous drones that will fly humans will most certainly want to know how to make their drones safer for the humans who’ll be riding in them. And thanks to MIT they just might have got their answer.

  3. Drones have been getting smarter and smarter lately, but that comes with a price point. I would love to buy one and fly it around, but it’s too pricy for me to justify the purchase. I tried a Phantom 4 once and it’s really fun.
    Hopefully in the future they will get a little more accessible, it would help my videos a lot. 😛

  4. This is really interesting, I have never used a drone before but I think that one of the most common problems is that they can’t really avoid objects which can lead to kind of expensive accidents, and this problem is ended with this drone function.
    And it actually looks surrealist, it looks like an spy bird from an action movie, lol.
    Thank you so much for sharing!

Leave a Reply