Quantcast

The Mechanics of Modern Autonomous Quadrocopters

By Wesley Fenlon

Quadrocopters are accessibly priced because the technology behind them is remarkably simple compared to the RC planes of a decade ago, but we've barely cracked the surface of autonomous control.

Quadrocopters like the Parrot AR.Drone and the smarter autonomous flyers represent one of technology's most exciting sectors. They're showing up at TED talks and Maker Faires. Former Wired editor-in-chief Chris Anderson recently left his position to focus full-time on his drone company. Drones have made RC and automated miniature aircraft far more affordable and open to the DIY community than ever before, but we rarely stop to think about how they actaully work.

Ars Technica has done just that, digging into the tech at work in every autonomous drone. While the flying machines are obviously complex, part of the reason they've been so successful is that, mechanically, their four motors are much simpler to control than a traditional helicopter rotor. Controlling the speed of each quadrocopter propeller determines the vehicle's tilt, altitude, and so on. Rotation is as simple as increasing speed of one rotor and decreasing speed of another.

Okay, that doesn't actually sound like it would be easy to control manually. But autonomous drones have microcontrollers specifically dedicated to the task and control algorithms that help it hang steady. A more powerful processor handles other tasks, like planning the drone's flight and relaying information back to home base.

Autonomous quadrotors are impressive flying all by themselves, but Ars' description of autonomous formations reveals how advanced these machines will become in the future:

Once the quadrotors can fly in formation, the path planning can be scaled up dramatically. In the nano-quadrotor video, each formation is treated like a single vehicle, with planning done by a ground computer. The system plans a path for each formation, treating it as if it were a single aircraft and communicating the path to the quadrotors. Each quadrotor knows where the formation should go, and they fly their formation-tracking along with it. So now, instead of having to compute trajectories for 16 quadrotors, the computer only has to compute trajectories for four formations. The quadrotors take care of themselves.

Control algorithms and sensors have to get better at predicting and reacting to wind and other complications. But they're already improving in other areas, like using 3D mapping to determine real-world location. Check out the full piece if you're interested in a deeper look at quadrocopter mechanics and programming.