Been a while since we’ve made an update, so I figured finishing our qualification video and technical paper was a good time to share. We’ve been making lots of good progress towards beating Mission 7. First, links:
Qualification Video: https://youtu.be/-DUE40ba2jw
Techical Paper: https://github.com/Texas-Aerial-Robotics/General/blob/master/IARC_Paper_2018/Paper.pdf
Now, in our qualification video, you can see our quadcopter autonomously takeoff. It then waits for a Roomba to pass in the camera view, which it is able to identify using our computer vision nodes. The Roomba positions are handed off to strategy node which calculates linear regression on the Roomba path, allowing us to predict where the Roomba will be 10 seconds in the future. This is obviously not perfect as the Roomba is coded to rotate a random number of degrees every 5 seconds, but it gets us to a position where we can block the Roombas forward path to direct it a different way. After linear regression spits out a waypoint in the correct viewframe, controls node takes that waypoint and performs a sanity check. If everything is within bounds, that waypoint is passed to the Pixhawk to fly the drone to that waypoint. After ariving at that waypoint, the drone switches to Land mode and it lands in front of the Roomba targetted.
We also would appreciate if you read through our technical paper, which is also linked above. This has all the information on the things we worked on to complete Mission 7.