Autonomous UAV Racing

An Intel research drone sitting in the indoor flight cage.

In the summer of 2019, I attended Beaver Works’ Summer Institute (BWSI) for a free six-week course on programming autonomous drones. While there, I attended seminars hosted by experts from Raytheon, BAE Systems, Lockheed Martin, and MIT’s labs. Throughout the course, we used ROS to utilize an Intel research drone’s forward-facing camera and downward-facing camera. The downward-facing camera identified an LED strip the drone would follow the path of LEDs. The forward-facing camera identified ArUco markers and the drone would fly over or under the marker depending on the fiducial’s ID. In the final two weeks, three other teammates and I worked on a program that would incorporate both ArUco and LED detection to complete an obstacle course as fast as possible.

Me and a friend being safe.

Me pondering the existence of drones.

We competed against several other teams and used PID control to control how fast the drone would navigate around obstacles. Initially, our focus was on completing the course. After some trial and error, we were the only team that was able to complete the course and held first place. However, another team was able to surpass our time by several seconds. In an attempt to regain our first-place position, we attempted to be more aggressive with our drone control and increased gains in our control loops. But, the drone would pitch forward too far and the bottom-facing camera lost sight of the LED track, making the drone failsafe. We were not able to regain first place, but we left the competition in second.

FPV Drones - NUAV

Fiber System Testing - Desktop Metal