Supervised learning tools applied to aircraft wake detection and formation flight

Everybody has already seen some geese flying in close V-formation. But did you know that they are actually saving energy which allows them to fly across much more long distances ? Although this technique improves a lot the performances of aircraft, it remains very dangerous because of the hard downwash at the center of the wake. Therefore the safety and the control of wakes are widely studied which comes up to the question of observability. How is a wake generated and how does it work ? Is a wake always observable ? Which onboard measures are useful in this task ? Wakes may be characterized as two big vortices generated at the wingtips of an aircraft. The wake thus generated will modify the wind speeds in the environment which will affect any aircraft encountering it. It will actually have an influence on its aerodynamic efforts which will give valuable information. However their complexity and the high non-linearity of the system avoids any analytical solution to be found. The path of supervised learning tools are then studied. In that respect, we will use neural networks and observe their performances for observing a wake. Supervised learning tools' ability to fit any function is well known and the wake of an aircraft makes no exception. They show that they are able to understand the high non-linearity of aerodynamical efforts induced by the wake of an aircraft. Besides, there remain some areas where they lack in performances but these are finally understood as being zones of unobservability of the system.