Swarming of Unmanned Aerial Vehicles by Sharing Distributed Observations of Workspace

A control and relative localization approach for a swarm of unmanned aerial vehicles (UAVs) flying in a forest environment is proposed in this paper. To achieve robust mutual relative localization of agents in such an obstacle-rich environment, we propose a decentralized localization approach based on a comparison of the workspace observation by on-board sensors of cooperating UAVs. We propose sharing sparse local obstacle maps to estimate bearing and distance between swarm members by fitting spacialy and time-distributed scans. Moreover, we propose fully decentralized flocking control rules adapted for deployment in such demanding conditions of real forests. The proposed approach was verified in the realistic Gazebo simulator, as well as in outdoor experiments. The approach introduced in this paper was also compared with a state-of-the-art method for relative localization and navigation of a swarm through a forest.