Science News

MIT’s Self-Flying Drone Can Avoid Crashes

Andrew Barry, a phD student and researcher at MIT’s Computer Science and Artificial Intelligence Lab (or CSAIL) has developed a system which allows a drone to detect obstacles and autonomously avoid them in a tree-filled field, all while travelling at 30 miles per hour.

The team working on the project are making their algorithm open source, according to the BBC, which means others who are interested in creating a self-flying drone of their own will soon be able to do so.

Everyone is building drones these days, but nobody knows how to get them to stop running into things.  Sensors like lidar are too heavy to put on small aircraft, and creating maps of the environment in advance isn’t practical. If we want drones that can fly quickly and navigate in the real world, we need better, faster algorithms.

Barry created his algorithm as part of his thesis with professor Russ Tedrake.  It runs 20 times faster than existing software, and it lets the drone detect objects and create a complete, real-time map of its surroundings.  The software is open-source and available online.  It operates at 120 frames per second and gives out depth information at a rate of 8.3 milliseconds per frame.

The drone he used weighs just over one pound and has a 34-inch wingspan.  It was made from a combination of common components costing about $1,700, including two small processors that could be used in a cell phone and a camera on each wing.

Past algorithms that tried to deal with the conundrum  of weaving past obstacles without crashing would capture images from each camera and then use them to try to search through the depth field for obstacles.  That approach was computationally intense, meaning that the drone was unable to move faster than 5 or 6 miles per hour

Barry’s realized that his drones could move at impressive speeds and the world doesn’t change much between frames.  Because of this realization, he was able to compute a small subset of measurements at distances of 10 meters away.

“You don’t have to know about anything that’s closer or further than that,” Barry explained.  “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.”

Click to comment
To Top

Hi - We Would Love To Keep In Touch

If you liked this article then please consider joing our mailing list to receive the latest news, updates and opportunities from our team.

We don't want an impostor using your email address so please look for an email from us and click the link to confirm your email address.