Game of Drones: Parrot Using Our Jetson TK1 DevKit to Show What’s Next for Drones

by Gary Rainville

Drone demos are everywhere at the International Computer Electronics Show this year. And a big part of the credit for this trend belongs to Parrot. The French company was the first to bring sophisticated drones that are controlled by smartphones to the mass market — rocking CES with its drone demo five years ago.

The key to Parrot’s mass market success: slick engineering. So it’s worth paying attention to where Parrot is going next. Besides its hot-selling AR.Drone 2.0 and Bebop Drone, Parrot is now showing how next-generation drones will rely on advanced computer vision capabilities to understand the environments through which they fly.

Pretty fly: Parrot's working with our Jetson TK1 devkit to explore where drones are going next.
Pretty fly: Parrot’s working with our Jetson TK1 devkit to explore where drones are going next.

The demo — dubbed “Kalamos” — uses NVIDIA technology to show a concept that’s key to next-generation drones. Unlike many other drone demos, Kalamos doesn’t rely on the satellite-based Global Positioning System (GPS), which can help human pilots navigate from point A to point B. Nor is this just about obstacle detection — a skill needed by drones so that they can avoid getting stuck in a tree, like Charlie Brown’s kite.

This is about building digital brains perceptive enough to model the environments through which they move.

To keep up with all that we’re up to at CES, visit nvidia.com/CES2015.

That’s no easy job. But Parrot’s Kalamos demo uses a simple model to show how it’s attacking this mind-bogglingly complex challenge. At the center of the demo: a two-foot tall playhouse complete with a garden in front and a bathtub on the second floor.

Staring down at this scene: dual cameras attached to our NVIDIA Jetson TK1 development kit, powered by a Tegra K1 processor. The devkit, in turn, is attached to a robotic arm that mimics the movement of a drone. An NVIDIA SHIELD controller directs the drone as its stereoscopic eyes do much more than just obstacle detection. They reconstruct the surfaces within the toy house in real time.

Using the parallel processing capabilities of the Tegra K1’s 192-core GPU, Parrot algorithms map out the scene in full color and display it on a nearby tablet. The resulting image shows accurate depth of field as the drone moves about the object, with precision closing in on one centimeter.

The potential applications for this are easy to imagine, and too many to list. But it’s clear already: the next generation of drones will be much more than mere flying machines. They’ll become tools that will help us better understand the environments we send them to explore for us.

For more on what we’re up to at CES 2015, go to nvidia.com/CES2015.