High-Speed Free-Flight Wind Tunnel Experiment With A Compromised Quadrotor

Creating A Dataset Suitable For Identification Of Aerodynamic Forces and Moments

More Info
expand_more

Abstract

To aid in the continued effort of making unmanned flight safer, this paper presents the experimenting effort towards determining aerodynamic forces on a off-the-shelf quad-rotor under compromised circumstances. For the first time an Incremented Nonlinear Dynamic Inversion controller is used for compromised flight in wind tunnel conditions. For the Parrot Bebop v1, five aerodynamically different configurations were tested. These configurations include reduced rotor effectiveness on one or two rotors and the inclusion of one or two bumpers. In order to gather a suitable dataset for these highly dynamic models, free-flight model identification is deemed necessary. To facilitate free-flight aerodynamic model identification, data was provided by the on-board inertial sensors and a motion-tracking OptiTrack system. Sensor fusion through Extended Kalman Filtering was chosen to counter frame vibrations and sensor bias. A sufficient convergence rate was found for estimating the biases. Further modeling efforts are required to validate these results. A rotor rate Kalman Filter was also designed to improve rotor rate differentiation. Excitation of the system was performed with position-controlled doublets, designed to observe theoretical modes from existing quadrotor models. This resulted in five main types of excitation, designed to maximize yawing motions, thrust variation and drag effects. The differences seen for each configuration show interesting behavior, and further investigation is recommended. The resulting dataset is considered uniquely suitable for estimating coupled dynamics with saturated actuators and loss of yaw control.