Aerial Robots

AscTec’s Pelican

This medium-size quadrotor [Fig. 1] is being used both indoors and out- doors for developing and testing 3D localization and mapping algorithms. Its navigation payload includes two on-board Inertial Measurement Units (IMUs), a GPS receiver, a camera for navigation and mapping purposes, and an on-board computer.

Quadcopter on a table with a lightweight cream-colored guard around its propellers and a tall stack of electronics in the center
Underside of a drone: a camera lens mounted below a stack of electronics between black carbon-fiber plates
Figure 1: (a) Ascending Technologies “Pelican” with on-board camera, IMU, and processor for performing navigation tasks. (b) View of the quadrotor payload.

DJI’s Inspire 1

This large-size quadrotor [Fig. 2(a)] is used outdoors for developing and testing 3D localization and mapping algorithms. Besides the sensors that the Inspire 1 comes with (IMU, GPS, and 4K camera used for stabilization, waypoint navigation, and aerial photography, respectively), we have instrumented this quadrotor to also carry a Velodyne PUCK 3D lidar, a stereo camera pair, a miniature-size IMU, and an Odroid ARM-based processing board for testing and validating large-scale localization and mapping algorithms [Fig. 2(b)].

White and black quadcopter with a camera on a gimbal hanging below its body
White quadcopter with black carbon-fiber arms and a blue mount underneath holding camera lenses
Figure 2: (a) DJI’s “Inspire 1” (b) “Inspire 1” equipped with additional sensors for large-scale localization and mapping.

DJI’s F330 and F400

These quadrotor frames are used for testing 3D localization, mapping, and navigation algorithms indoors. The F330 [Fig. 3(a)] is ideal for flying with a small-size IMU, and one or two cameras in different configurations while using an ODroid processing board for supporting its onboard vision-aided inertial navigation system (VINS). The F450 [Fig. 3(b)] has larger payload capacity and can carry Google’s Project Tango cell phone, which has a fisheye camera and a quad-core ARM processor. Both quadrotors are also equipped with PX4Flow sensors that comprise a narrow field-of-view (fov) camera and a ultrasonic range sensor for measuring the speed and distance from the ground of the vehicle. Lastly, these two quadrotors carry sonar rings used for obstacle detection and avoidance.

Side view of a quadcopter with red and white arms, four DJI motors, long landing legs, and a camera on a blue mount underneath
Quadcopter with labeled parts: NAZA controller on top, optical flow sensor underneath, eight ultrasonic range finders, and a cell phone camera
Figure 3: DJI’s “F330” (a) and “F450” (b) frames modified for autonomous navigation.

Parrot’s Bebop

This quadrotor [Fig. 4(a)] is the new, upgraded version of the Parrot’s ARDrone. It was designed for teleoperation using a mobile device, and as it predecessor, it includes two cameras (forward and downward facing), an IMU, a GPS receiver for outdoor navigation, an ultrasonic range sensor for maintaining a constant height when flying close to the ground, and a wireless router for communication. One of its main advantages, as compared to the ARDrone, is the increase in processing power. The Bebop carries a custom-made processing board with a dual-core 800 MHz ARM processor, which we use for implementing and testing autonomous navigation algorithms based on visual servoing and VINS.

Red and black Parrot Bebop quadcopter with four propellers and curved bumpers
Gray Parrot AR.Drone quadcopter with a foam guard ring around each of its four propellers
Figure 4: (a) Parrot’s “Bebop” (b) Parrot’s “ARDrone”.