Aerodynamic Modelling and Wind Disturbance Rejection of Multirotor Unmanned Aerial Vehicles

Abstract

The use of multirotor unmanned aerial vehicles (UAVs) for applications close to the environment, such as physical sampling, inspection, and navigation in narrow environments, has increased drastically in recent years. Wind disturbances both increase the risks of collision and decrease the precision of the work performed by UAVs and are, therefore, a major limiting factor for these applications. This work aims to improve UAVs’ wind disturbance rejection performance by investigating the robust control of a canted-rotor octorotor capable of vectored thrust. A UAV dynamics model inclusive of aerodynamic forces and moments is first developed for use in robust control synthesis. Aerodynamic polars are fitted to static load-cell data and validated with free-flight station-keeping experiments in a wind tunnel. The model is found to predict the experimental root-mean-square (RMS) position error along the direction of the wind within 7%. A new vectored-thrust controller is then developed based on this model. It comprises a motor mixer converting desired torques and vectored thrust into motor commands, an off-the-shelf attitude controller, and a novel dynamic output feedback H-infinity position controller. Frequency-dependent weighting is applied to use attitude control to reject low-frequency disturbances and vectored thrust for high-frequency disturbances. Comparisons to PX4 Autopilot, a widely used baseline flight controller, are made both in simulation and in free-flight experiments in a wind tunnel at wind speeds up to 12.8 m/s. The H-infinity controller is found to halve the RMS position errors in most cases, at the cost of increased actuator usage. Finally, an investigation of the benefits of wind velocity feedback for station-keeping is conducted in simulation. Two feedforward pitch controllers are created and integrated with an existing controller, showing promising station-keeping results with RMS position errors up to 66% lower.