Aerodynamic Disturbance Estimation in Quadrotor Landing on Moving Platform via Noise Reduction Extended Disturbance Observer

Abstract

As the main factor affecting the safety of quadrotor unmanned aerial vehicles (UAVs) on moving platforms, aerodynamic disturbances are not easy to directly measure but can be effectively estimated from control system information by extended disturbance observers (EDOs). To guarantee estimation accuracy for aerodynamic disturbances with fast dynamics induced by increased speed of landing platforms, high bandwidth is necessary for EDOs. However, high bandwidth of EDOs will result in high gain problems which may amplify measurement noises in the control system. To suppress the effects of measurement noises on estimation accuracy, a pair of noise reduction EDOs (NREDOs) are proposed to estimate aerodynamic disturbances for quadrotor UAVs landing on moving platforms. The pair observers are designed to estimate force and torque disturbances for translational and rotational subsystems, respectively. Different from EDOs, each NREDO takes the integral of the lumped disturbance as an augmented state and virtual measurement in the state-space disturbance model. The prediction error of the virtual measurement is taken as an innovation to update the observer. Moreover, a tuning rule of observer gains is proposed to further improve estimation accuracy. Theoretical analysis indicates that the integrals provide NREDOs with superior performance in noise suppression than EDOs. Landing experiments on a platform of 25 km/h demonstrate the effectiveness of the proposed scheme.