Setup and configuration of an autonomous UAV swarm for indoor coverage path planning

Nowadays, there is excitement among professionals over the incoming robotics technologies that are paving the way to make a more efficient, sustainable and comfortable working environment. In particular, in the last decades the UAVs (Unmanned Aerial Vehicles) are being exploited in several civil sectors such as agriculture, photography/videography, construction, environmental monitoring and even transport. This list can potentially grow further, thanks to the intrinsic versatility of drones and their capability of operating autonomously. However, the bottleneck of these devices relies on the methods used to localize them in space. Outdoor, satellite systems like GNSS (Global Navigation Satellite System) and GPS (Global Positioning System) allow drone localization with acceptable precision. Nevertheless, this consideration cannot be applied to indoor spaces, where It isn’t possible to rely on such systems to get an accurate estimate of the drone position. Hence, the goal of this master thesis project is to configure an autonomous mission for a swarm of drones based on VICON localization system, while using state-of-the-art software such as ROS2 Humble and PX4 Autopilot v1.14. Specifically, the tests were conducted inside a cage of size 7x3x3 meters located in the Collaborative and Service Robotics Lab at LINKS Foundation. In the beginning, several experiments were performed to try and configure correctly the drones to fly. This thesis reports in detail how the setup of both hardware and software was conducted, highlighting the problems faced and the solutions found. The next step was to enable a fleet of multiple drones to perform an autonomous mission, to test the estimate system and the configuration done. The chosen task was mCPP (Multi Drone Coverage Path Planning) using existing algorithms: a STC (Spanning Tree Coverage) algorithm and DARP (Divide Areas based on Robot’s initial Positions) algorithm. Moreover, the mCPP task was an offline one, meaning the environment in which the drones executed the mission was fully a priori known. A detailed analysis of the state-of-the-art of such a task and on how the above mentioned algorithms were implemented, along with the performance results is shown in this document. At the end, some considerations on further work and suggestions for future experimentation will be given.