If you're seeing this message, it means we're having trouble loading external resources on our website.

If you're behind a web filter, please make sure that the domains *.kastatic.org and *.kasandbox.org are unblocked.

Main content
Current time:0:00Total duration:5:37

A crash course on indoor flying robots

Video transcript

[MUSIC PLAYING] This is a quadrotor. It's called a quadrotor because it has four propellers that spin and generate thrust. More on that in a second. This is the pilot controlling the vehicle with a radio transmitter. That's pretty neat. But if we take a short trip across the street-- of course looking both ways before we cross-- we come to a place where this quadrotor can fly by itself, without any human help at all. We don't even need a pilot. This flying robot can operate with extreme precision in tight indoor spaces, and can do some other pretty neat stuff as well. So if you're wondering how to make robots fly, you've come to the right place. [MUSIC PLAYING] Maybe crash course isn't the right term. [MUSIC PLAYING] To figure out how to make robots fly, we'll need to understand the basic physics of quadrotors, how humans pilot them, how we can use a computer to achieve the same task, and why the resulting flying robots can do more complex things. First, let's take a quick look at the physics behind how the quadrotor flies. When the propellers spin, they push downward on the air around them. Newton's third law tells us that the air applies an equal and opposite reaction force on the propeller. When this lifting force equals that of gravity, the quadrotor achieves hover flight. In order to bank, one propeller spins slightly faster than the opposite one. This introduces a horizontal force, in addition to the one opposing gravity. And the vehicle moves sideways. That's great. But it doesn't tell us how the quadrotor can rotate about its vertical axis. It turns out that Newton's third law also applies to rotational force, called torque. When these two propellers spin, they apply a torque to the air in the clockwise direction. The air applies an equal and opposite reaction torque, pushing the vehicle in a counterclockwise direction. Meanwhile, the other two motors spin in the other direction, plus the reaction torque pushes the vehicle clockwise. When all four motors are turned on, the rotational forces-- remember they're called torques-- balance each other. In flight, the vehicle turns by spinning two motors ever so slightly faster than the other two. Now we know the basic physics of how a quadrotor flies, but before we can make it fly robotically, we need to know how to control it. First, let's figure out a human would do this. The task can be broken down into four key steps. First, the pilot uses his eyes to observe the vehicle and figure out where it is, and in which direction it's pointing. In this example, let's say that the pilot sees that the quadrotor is sinking. Next, the pilot has to decide what control commands to give the vehicle. In this case, the pilot has to stop the vehicle from sinking, and thus decides to increase the speed of all four propellers. To tell the quadrotor what he's decided on, the pilot uses a radio transmitter, which is basically a fancy remote control. Finally, the quadrotor listens for the radio commands, and adjusts the speed of each motor accordingly. Now let's see how each of these four steps changes in order to make the quadrotor fly robotically. In the first step, we use specialized cameras for vision, instead of the pilot's eyes. The cameras shine infrared light, which bounces off of small reflective markers on the vehicle, and go back to the camera. A camera from this side point of view can tell how far the marker is in the vertical direction, and one horizontal direction. And a camera from this top point of view can tell how far the marker is in both horizontal directions. Using some slightly more complicated mathematics, we can use the points of view from 12 different cameras mounted along the ceiling to determine the exact three-dimensional position of the markers. This process is executed many times per second to check the position of the markers, and plus the quadrotor, in real time. In step two, we use a computer to calculate the control commands, instead of the pilot's brain. The computer program consists of a couple hundred lines of C++ code, written by grade students who really don't get out much. It does essentially the same thing as the pilot, using the observed position of the quadrotor to calculate control commands, only it does so in a much faster and less dramatic fashion. In step three, the system uses a similar radio transmitter, except a smaller one without any switches or control sticks. Step four is exactly the same as before. The quadrotor listens for radio commands, and adjusts the speed of each motor accordingly. So we've updated all four steps in order to make the quadrotor to fly entirely by itself. Now all we need is for our grad student to press the go button, and voila. One of the reasons the robots fly more precisely than the human pilot is because this loop of information-- called a feedback control loop-- can be executed much more quickly by computers. In this case, 200 times per second. This allows researchers to do cool things with these indoor flying robots. For instance, fly six of them at once. Or why not 10? They can teach the vehicles how to switch out their old batteries for new ones automatically. Or stop a payload from swinging. They can even do flips like this one. Or this one. Or this one. And the fun doesn't stop with quadrotors. The same technology can be applied to weirdly shaped three-winged maneuvers. Or more conventional fixed-wing vehicles like this one, this one, and this one that can even fly loops. Well hopefully you've learned the basics of how to make robots fly. This concludes the crash course-- I mean, expedited learning experience. [MUSIC PLAYING]