I’m sure you’ve had this conversation at one time or another. “These planes they just fly themselves! The pilot sits back and lets the autopilot do all the work.” I’ll just refer readers to Patrick Smith’s long running “Ask the Pilot” blog for the best rebuttal when it comes to human sized aircraft flying themselves.
Drones are a little different, especially quadcopters. A quadcopter is always adjusting motor speed to fine tune its position. Watching a quad on a windy day you can clearly see how much it has to work to maintain a stable position. But when you look at the video feed it is rock steady, and the pilot might not be controlling the drone at all.
A pilot sends commands to the drone using the controller. Controllers typically have a pair of joysticks, a number of buttons and/or switches and either an integrated display or a holder for a mobile phone or tablet that serves the same purpose. During manual flight the pilot uses the joysticks for control of pitch, yaw, roll and throttle to move the drone through the air, but doesn’t directly control the speed of the motors. The autopilot translates the input signals to motor speed controls. If the pilot wants the drone to move forward, he will push the right joystick up (or away). The autopilot receives the message and will increase the speed of the rear motors and/or decrease the speed of the front motors, causing the drone to pitch down, and then increase the speed of all the motors to cause forward movement. When the drone reaches the new destination the pilot will release the joystick, allowing it to center, and the drone will adjust the motor speeds to level itself and hold the position. If the pilot then wants to aim the camera, he can then push the left joystick to the left or right. The autopilot will then change the speed of two of the motors on opposite corners (left-front and right-rear, for example) to change the rotational balance, which will cause the drone to yaw.
Why does this matter? Well, because early quadcopters were extremely hard to fly. The pilot had to constantly monitor the aircraft and make adjustments to keep it in the air. Racing quads still are flown this way. But for most drones, the key to stable flight is the inertial measurement unit (IMU) and Global Navigation Satellite System (GNSS) receiver.
The IMU provides feedback to the autopilot so it can keep itself level. If the aircraft pitches forward, the IMU will indicate this and the autopilot can adjust the motors to compensate. If the pilot sends a roll starboard command the IMU will limit the roll so that the drone doesn’t flip over (this can be ignored in FPV drones for interesting effects).
What the IMU cannot do is know what the air is doing. Much like a rowboat on a river, if the wind is blowing the drone will drift along with it. The way to keep a drone in one spot is with a GNSS receiver.
The GNSS is a high performance GPS (US), Galileo (European) and GLONASS (Russian) receiver. It is able to average all three (and increasingly the Chinese BeiDou constellation), in order to get a more accurate position than possible using one system alone. It is able to update the autopilot 8 times per second. Some higher end receivers are able to supply updates at 15 Hz or more. For comparison your phone’s GNSS chip might update once or twice a second, and the operating system may que data adding more delay. Drones that are used for mapping and other high precision use will use two and average them, along with a ground station that can provide corrections for atmospheric conditions and other errors.
The idea is the same as the IMU: provide feedback to the autopilot to maintain position. The more accurate the postion, the more stable the flight. Most drones will not take off until the GNSS receiver is locked and has a good location. If there’s interference or issue that causes the GNSS receiver to lose signal, the drone can become unstable and crash. But that doesn’t happen very often if the pilot isn’t performing a risky maneuver such as flying under a metal structure.
So the autopilot can keep the drone stable, will respond to inputs from the pilot and knows where it is on Earth. It isn’t all that difficult to imagine automating a flight, and that’s exactly what many commerical pilots do. Flight planning software will create a mission, from takeoff to landing with a few mouse clicks.
This can be a powerful tool for mapping the world, tracking construction progress over time, building 3D models or any sort of survey work. The drone can fly the path far more accurately than a human. Not only that, but because every action is time and location stamped in the flight log, any data gathered (photos for example) can be tagged with their precise location, which makes it possible to build extremely accurate maps and 3D “digital twin” models that can be accurate to a few millimeters.
Another form of autonomous flight is programming keyframes. Instead of having to repeatedly make complicated controler inputs, the pilot can set up keyframes and have the drone fly between them automatically. My Skydio 2 drone does this very well, and uses artificial intelligence to make the moves smoothly, while actively avoiding stationary objects.
The other function that is somewhat unique to Skydio is its abililty to track a person or vehicle while avoiding hazards.
So yes, once you get the hang of it, flying drones is pretty easy. In part 2 I’ll get into just how hard it is to be a drone pilot.