Wind Estimation for Small Fixed-Wing UAV Using Standard Sensors

Small Fixed-Wing Unmanned Aerial Vehicles (FWUAVs) are very sensitive to wind since its magnitude is comparable to the FWUAV speed. Moreover, their small size and the limitation of their propulsion system and actuation capability make it difficult to maintain control in windy conditions. In order to limit wind impact in FWUAV attitude and trajectory, instantaneous wind velocity estimation can be introduced in control laws. When aircraft is equipped to measure the three components of the
more » ... vector, typically either with a multi-hole Pitot probe, or a simple Pitot probe and vanes to measure angle of attack (AOA) and angle of sideslip (SSA), wind speed may be computed directly or estimated using geometric relations between ground speed, airspeed and wind speed. However, unlike larger aircrafts, small FWUAVs do not usually have sensors dedicated to measurement of these variables due to difficulties encountered in their installation, their cost and to aerodynamic flow disturbances nearby fuselage. Standard UAV sensor suite contains GNSS (Global Navigation Satellite System), IMU (Inertial Measurement Unit), and Pitot tube. Another approach, frequently used to estimate SSA on board transport aircrafts, consists in replacing SSA measurement by its estimate obtained from aerodynamic modeling, Pitot tube and IMU measurements. But this approach is based on a reliable aerodynamic model, which may be tuned by wind tunnel or flight tests. Such models are seldom available for small UAV applications. Within the framework of wind estimation using small FWUAV [1], we propose an approach which requires neither AOA and SSA sensors nor high-fidelity aerodynamic models but is based on measurements provided by a standard sensor suite and simplified aerodynamic models used as virtual AOA and SSA sensors. Virtual sensors models are based on relationships between data coming from Pitot tube, accelerations and actuators deflections, and some aerodynamic parameters which need to be estimated together with a scaling factor which calibr [...]
doi:10.13009/eucass2022-7135 fatcat:zvxepxnc7vgo7ksa2ixwxcpi3u