Parrot’s Ubuntu- and ROS-driven, Tegra K1 based “S.L.A.M.dunk” development kit has a stereo camera and sensors that adds obstacle avoidance to drones.
Parrot’s S.L.A.M.dunk, which is named for its integrated Simultaneous Localization and Mapping algorithm, can be added to any Linux-driven drone to help it navigate indoors or in other barrier-rich outdoor environments where GPS signals are not available. Assuming obstacle avoidance technology can be sufficiently refined, indoor package delivery may be the next big application for Unmanned Aerial Vehicles (UAVs). There are also many outdoor environments such as wooded and urban areas where drones struggle to navigate by GPS and standard imaging alone.
Parrot’s S.L.A.M.dunk camera and sensor unit
(click image to enlarge)
Due to ship in the fourth quarter, the Parrot S.L.A.M.dunk provides a sensor array and “open” development kit to enable simpler drones with more autonomous functionality, says Parrot. The French firm is known for its Linux-based prosumer drones, such as the AR.Drone, BeBop 2, and soon to ship fixed-wing Parrot Disco. Last week, it also announced two more Linux-based consumer drones: the Mambo and Swing (see farther below).
S.L.A.M.dunk mounted on a UAV
(click image to enlarge)
The Parrot S.L.A.M.dunk’s stereo camera and sensors can also be used for 3D mapping and basic data gathering. In addition to supporting quadcopters, it supports flying wing drones, as well as articulated arm and mobile robots. It can even be used to prototype automated driving, says Parrot.
The 140g (4.9 oz) device runs Ubuntu 14.04 and ROS (Robot Operating System) on an Nvidia Tegra K1 system-on-chip. The quad-core, 2.3GHz Cortex-A15 SoC integrates adds a 192-core Kepler GPU supported by Nvidia’s CUDA graphics libraries, enabling computer vision and deep learning in addition to display rendering.
The S.L.A.M.dunk integrates a fish-eye stereo camera that offers 1500 x 1500-pixel resolution at 60fps. There are also onboard inertial-measurement unit (IMU), ultrasound, magnetometer, and barometer sensors. The device can hook up to a drone’s power source via a 3.5mm jack and micro-USB port, and is claimed to minimally impact battery life.
The S.L.A.M.dunk also provides a micro-HDMI port. “Being able to run an Ubuntu desktop directly from the device is a great way for developers to do quick iterative development directly on the board and test their results literally on the fly,” says Parrot.
An Ubuntu/ROS combination is the most common flight stack among Linux drones these days, including models that conform to the Dronecode standard. There was no specific mention of Dronecode here, but Parrot is a member of the project and is aligning its technology with the platform.
Qualcomm offers its Ubuntu/ROS based Snapdragon Flight development platform running on a Snapdragon 801 SoC. This week, the chipmaker announced plans to test Snapdragon Flight based UAVs controlled over AT&T’s 4G LTE and emerging 5G cellular networks. The tests anticipate the possible loosening of FAA restrictions in specific applications that forbid operating drones beyond line of sight (and WiFi).
Meanwhile, Intel recently announced a somewhat similar Aero Compute Board and quadcopter, promising improved obstacle navigation based on its Intel RealSense camera. Instead of running Ubuntu, however, it uses a uses a Yocto Project based Linux stack running on an Intel Atom processor.
At last week’s IFA show, Parrot announced two more Linux-based consumer drones: the Parrot Mambo and Parrot Swing. The $120, indoor-ready Parrot Mambo mini quadcopter can carry attachments to catch objects and throw projectiles. The $140 Parrot Swing is an unusual, X-shaped fixed wing device that also offers autopilot operation and a vertical takeoff and landing mode.
The Parrot S.L.A.M.dunk will be available in the fourth quarter at an unstated price. More information should eventually appear on the Parrot website.