This video demonstrates the functionality of the color tracking code. The algorighm takes in a video, and can find and track the green circle. The algorithm uses the RGB image to find the center of the green circle, then maps that pixel to the depth image to find how far away the circle is.

With that information, we are able to send velocity commands to the Pixhawk to fly the drone in order to keep the center of the ball centered.

This is a video of the First Flight on April 9th 2017.

We are controlling the quadcopter with a Futaba TC7 controller communicating to the Pixhawk Mini through an SBus compatible RC receiver. We set the flight mode to stabilize and tested if the hardware setup of the drone was working and able to fly.

We are flying the drone in STABILIZE mode, which means we control the direction the drone flies, and it will keep the drone level with the ground.

Flight test using QGroundControl sofware, provided by the PX4 flight stack, to set the GPS waypoints. After it went to all the waypoints and went back to home, the flight mode is set to AutoLand to safely land the drone.

We first arm the controller in stabilize mode, then set it to mission. We preset the waypoints and home in the QGroundControl. To figure out what GPS coordinates we wanted to fly to, we stood in differet areas of Tech Green and noted down the GPS coordinate of that location.

In this test, we used telemetry to communicate to the Pixhawk with a python script. We set the flight mode to autotakeoff, wait for a few seconds, then set it to autoland.

We are currently controlling the odroid through SSH so we can see a log of what is going on and stop any program that could cause damage to the drone.

The Pixhawk is using a GPS lock to keep the drone from drifting side to side. As it lands, you can see that it drifts a little. This is due to the GPS lock not being very accurate because of all the trees and buildings in the test area.

To localize the quadcopter, we are using ORB SLAM. SLAM stands for Simultaneous Localization and Mapping. With this algorithm and the RGB-D images from the Realsense camera, we are able to extract features from the images and determine where the quadcopter is in the world.

In the video, we are moving the camera around the room. The algorithm is able to detect the features in the video, shown on the right, and mapping it to a point, shown on the left.

This is video captured from the R200 camera while it is in flight.

This is the final build of the quadcopter with all the components mounted. The resulting weight was 1.6 kg.

This video demonstrates the functionality of the ORB SLAM algorithm with video captured from walking around with the drone around the Van Leer Building.