Watch MIT’s speeding drone fly around trees all on its own

3 Nov 2015 | Author: | No comments yet »

Fast-Flying Drone Avoids Obstacles: Here’s How.

An MIT researcher has developed an autonomous flying drone capable of avoiding obstacles on its own while speeding up to 30 miles per hour, MIT said in a statement on Monday. Whether you’re for or against the rampant use of drones, you can probably agree that it’s best if unmanned aerial vehicles don’t hit people or things.

One of the biggest issues preventing drones delivering packages—other than many, many regulations—is figuring out how to make sure they don’t crash into things. In that vein, a Ph.D. candidate at MIT’s Computer Science and Artificial Intelligence Lab (CSAIL) recently demonstrated a new obstacle detection system that enables a drone flying at 30 mph to navigate through trees and other obstacles. Using Barry’s stereo-vision algorithm, the drone constantly builds a map of its surroundings and zips in and out of trees without any human control. As part of his thesis work with Professor Russ Tedrake, Andrew Barry developed a stereo-vision algorithm that detects objects and builds an accurate map of the surrounding locale in near real-time, according to MIT. If we want drones that can fly quickly and navigate in the real world, we need better, faster algorithms.” The results look like lost scenes from the Top Gun special effects sequences.

The open source software, which works at 120 frames per second and some 20 times faster than existing technology, is available for download at Github. The hard part is “how to get them to stop running into things.” Barry’s alternative is an algorithm that uses some clever and efficient programming to make a drone fully autonomous. The beauty of this technology is that it lessens the need to weigh down drones with heavy sensors, Barry, told MIT News. “Sensors like Lidar are too heavy to put on small aircraft, and creating maps of the environment in advance isn’t practical. The problem itself is fairly simple: Small-scale UAVs like the ones many amateurs and tinkerers own aren’t designed to autonomously avoid obstacles because they aren’t capable of carrying the weight of the processors they’d need to analyze the world around them and react to it. This is all interesting, but the obstacles in this case are stationary; I’m guessing a lot more work needs to be done so drones can also avoid birds, people, cars, and, errr, other drones.

There are currently a number of efforts to develop obstacle avoidance systems for drones, and Qualcomm has built a reference design for drones using standard components. This new system maps only 10 meters out, making it much less computationally intensive, although what happens if a squirrel suddenly enters the previously empty path between you and that 10-meter mark is not exactly clear. Rather than try to analyze every object in every frame captured by a drone’s camera, they set up a threshold distance–10 meters–and only analyzed those frames. Research maven Mary Meeker, partner at Kleiner Perkins Caufield Byers, said she expects the consumer drone market to grow 167% year-over-year, with worldwide shipments reaching 4.3 million units this year, netting total revenue of $1.7 billion by year’s end.

The goal of work like Barry’s is to mitigate the risk of using potentially very useful technology not just for package delivery but for building or land inspections, journalism, even fire fighting. (And if, despite all this great research, a drone does clock someone there’s now insurance for that.) But because it’s flying at 30mph, it doesn’t need to spend time analyzing all those extra frames and objects. “While this might seem limiting, our cameras are on a moving platform (in this case, an aircraft), so we can quickly recover the missing depth information by integrating our odometry and previous single-disparity results,” they write in a paper published on ArXiv.

Here you can write a commentary on the recording "Watch MIT’s speeding drone fly around trees all on its own".

* Required fields
Our partners
Follow us
Contact us
Our contacts

ICQ: 423360519

About this site