Multi-UAS Wildfire Mapping

In 2019, I competed in the Air Force Research Laboratory Swarm and Search AI Competition with two of my labmates and advisor. The goal was to use teams of UAS to find a map wildfires while avoiding complex terrain and burning in fires. I developed the planning module that relied on computationally efficient exploration and line following methods to achieve this safely. Jeremy Castagno developed the boundary estimation method that took in fire and free point detections in a convultional neural network framework to produce a polygonal fire estimate. We ended up placing first in the hackathon, winning a total of $26,000. After the competition we were able to publish our approach as a journal paper [1].

[1] J. Castagno, M. Romano, P. Kuevor, and E. Atkins, “Multi-unmanned-aerial-vehicle wildfire boundary estimation using a semantic segmentation neural network,” Journal of Aerospace Information Systems, 2021. Google Drive. ARC AIAA

UMich Article

We were honored with an award by the Sea Lord of the Royal Navy aboard HMS Queen Elizabeth (below). We are on the left along with our British counterparts (UK winners) on the right. The red, white, and blue dress shirts were a happy coincidence.