Vision based automatic landing
Last January, Airbus demonstrated the first fully automatic vision-based take-off. Then, a fully autonomous runway landing occurred as part of a flight to Beijing to deliver COVID-19 related medical equipment to hospitals in Europe.
This is the first time ever an aircraft has been able to take off taxi and land totally autonomously.
I trust that this recent collective success is partly due to the “Visioland” project that I had the pleasure to lead between 2013 and 2017. I'm also a co-author of a US patent on vision based automatic landing registered by Airbus in 2018.
The aim of Visioland (VISIOn-based aircraft LANDing techniques) was to develop vision based control solutions to perform the automatic landing of an aircraft (or a fixed wing drone) on an unequipped runway and in a GPS denied environment. With colleagues from the French Aerospace Lab, Airbus, Inria and the French CNRS, I worked on the following solutions
(1) observability analysis and observer design for perspective vision systems,
(2) novel estimation based or discontinuous anti-windup solutions,
(3) saturated vision based control laws analysis techniques
(4) applications of recent backstepping techniques to visual servoing including the development of controls under input saturation.
The interested reader will find all the details in my publications list.