Asynchronous Perception and Control on Quadrotors
Image credit: Comité JNRR
Abstract
Quadrotors play a crucial role in difficult tasks such as exploration, inspection and search-and-rescue in dark and cluttered environments. These are usually GPS-denied environments, and the positioning required for autonomous navigation of quadrotors must therefore be based on onboard visual inertial odometry. This embedded localization can quickly become computationally and energy intensive, and even prove ineffective if the complete perception and control loop is designed in a conventional, periodic manner. As an alternative, we propose rethinking the way quadrotors perceive and navigate their environment based on asynchronous strategies, which can be observed in many animals in nature.
Location
Rennes, France
IRISA - Université de Rennes - Campus Beaulieu, Rennes, Bretagne 35000
Authors
Doctor in Control and Robotics
My research focuses on the efficient navigation of UAVs in constrained environments. I recently joined the LEASARD CominLabs project at LS2N Nantes, in which I investigate SLAM and navigation algorithms with event-based cameras to improve drone autonomy in cluttered environments. This work is supervised by Isabelle Fantoni. My Ph.D. thesis work, led at ICube Strasbourg and GIPSA-lab Grenoble, has been dedicated to the event-driven navigation of a drone in dark environments, as a part of the dark-NAV ANR project. This thesis has been supervised by Nicolas Marchand and Sylvain Durand. Through my research work, I developed research interests in aerial robotics, robust control and set theory. As a graduate of an engineering school (INSA Strasbourg), I have developed a taste for control theory, which I try to deploy as much as possible on real control systems.