PhD Thesis on Machine Learning for Helicopter A
Airbus Helicopters CIFRE thesis of 3 years (fixed-term contract)
Machine Learning applied to helicopters flight parameters estimation: airspeed, angle of attack and sideslip on the whole flight envelope
Keywords: machine learning, estimation, airspeed, helicopter
The current technology installed on helicopters (Pitot probes) is not able to measure all the components of airspeed vector in the whole flight envelope. Only the longitudinal component along the helicopter body axis is measured. The lateral and vertical components measurements are not available on helicopter. Additionally the air data are greatly disturbed at low speed (<30kts, low sensitivity and noise from rotor downwash). This lack of information limits the maneuvers near hover (main interest of helicopter configuration) and in cruise where the pilot shall respect the flight envelope limitations.
Alternate solutions must be defined and could be based on the following elements:
At low speed, airspeed estimation could be used to compute the wind that will be used by the pilot to monitor the flight envelope or by flight control laws to enhance the tuning. At high speed, the sideslip estimation could be used to optimize the helicopter performance, to alleviate flight loads on structural elements or to enhance decoupling functions.
The aim of the thesis is to propose an estimator based on machine learning and to compare its performance with the state of the art.
The research will detail the following aspects:
(c) GdR 720 ISIS - CNRS - 2011-2018.