An expert and well-illustrated interview with the co-founder - intelligent and probing questions about its design and capabilities. The design goal was to make a drone that can be flown by a novice, who can actually concentrate on the activity they're doing instead of having to fly the drone.
This $2500 drone has 13 cameras, and autonomously avoids tight obstacles.
It's a long video (22') so here are where you can find the highlights.
At 7:13, he describes that the onboard computer is predicting a full 4 seconds ahead of your current position, to plan its route to follow you. It can also lead ahead of you, looking backwards, and still avoid objects and predict where you're going to go. This is best illustrated at 4:40, with a switchback on a running trail, and a large tree obstructing its path, yet it expertly, autonomously navigates and keeps the runner in frame.
Avoidance: He mentions at 9:34 that they experimented with many types of sensors like laser and sonar, but found that black and white cameras were the cheapest and most useful (light, cheap, and information-rich at 10:19.) The processing of 13 separate images in real time is nicely illustrated at 5:41 and again at 5:57.
The ability of portable devices to recognize objects has come become a thousand times faster in the last decade. This video illustrates that capability:
The image quality is a little grainy in this interview video, but it's better illustrated in the company video at 2:26 :
https://youtu.be/Vtlr3Q9JKUo
Here's a good demo of how skilfully the drone follows a skier down a canyon and avoids obstacles. Notice at 2:18 where it avoids an obstacle by going under it.
.
https://youtu.be/g_HURAlud1EHere's a good demo of how skilfully the drone follows a skier down a canyon and avoids obstacles. Notice at 2:18 where it avoids an obstacle by going under it.
.
No comments:
Post a Comment