Search and Rescue Under the Forest Canopy Using Multiple UAS

We present an autonomous system consisting of multiple unmanned aerial vehicles (UAVs) for search and rescue under the forest canopy. Our vehicles can be rapidly deployed, can collaboratively explore expanses of terrain efficiently, and are agile enough to operate in reasonably thick forests. We demonstrate the ability to carry out GPS-denied exploration with on-board pose estimation, map inference, and path planning. In addition, we utilize a place recognition system that is able to handle perceptual aliasing unique to a forest environment, and fuse individual areas explored by each vehicle into a globally consistent map. We perform extensive evaluations in both simulation and real-world to demonstrate the effectiveness of our proposed system.