These topic SCHWEITZER Thibault and theses.fr

Abstract :

A drone swarm is a coordinated set of drones with the aim of performing a common task. It takes advantage of cooperative behavior to broaden the scope of action and improve overall operational efficiency.
The localization of a swarm of drones in conditions without GPS remains a topical issue for an effective implementation of swarms of drones. In the literature, there are two techniques used, the first is to add beacons (visual, RF, RTK, UWB, Optitrack, ...) in an environment known a priori so that the drones position themselves in relation to these . However, the second technique only uses on-board sensors without beacons to position itself in an unknown environment. In this thesis, we will study several unknown environments (dark, metallic, with or without GPS coverage, ...) and the different modules (VIO, Lidar, UWB,...) used for locating a swarm of drones. In the literature, localization methods have already been developed, among which those that use the visual localization method. The method relies on visual localization modules on board all drones that provide an estimate of the relative positions of neighbors in the swarm and their positions. However, methods based on visual localization suffer from limitations mainly related to the limitation of the field of vision of the cameras and the luminosity. Another method using distance measurements from Ultra Wide Band (UWB) transceivers can be used to calculate drone positions. However, drift can be significant for long trajectories, especially when the environment is visually challenging. Using data fusion, several decentralized systematic solutions have been proposed for the autonomous navigation of a swarm of drones in unknown scenes with multiple obstacles using only on-board resources. Recently, it has been reported that the decentralized visual-inertial-UWB fusion allows the estimation of the relative state of drones with an accuracy of the order of centimeters.
This thesis joins these efforts for the development of decentralized robust localization systems for the navigation of a swarm of fully autonomous drones in environments (indoor and outdoor) with a high density of obstacles. The objective of this thesis is to study the limits of each localization module with respect to different complex environments and to propose an intelligent and robust localization method in the face of the complexity of the environment to ensure operational safety. of the navigation system of a swarm of drones.

Supervison :

Under supervison of Associate Professor Moncef HAMMADI (ISAE-Supméca)