Compensating Aerodynamics of Over-Actuated Multi-Rotor Aerial Platform With Data-Driven Iterative Learning Control

The performance of over-actuated Unmanned Aerial Vehicles (UAVs) is significantly influenced by aerodynamic effects, as they are subject to different airflow conditions. For those aerodynamic effects that are difficult to model but could occur repeatedly, Iterative Learning Control (ILC) has great potential to improve the system performance by learning from the error of previous trials. In this letter, we studied and implemented the performance of the data-driven ILC algorithm on the over-actuated UAV platforms to compensate for the unknown aerodynamic effects. To facilitate the design and implementation of ILC for the nonlinear dynamics of the over-actuated UAV platform, we encapsulated the control system to create a 6 Degree-of-freedom (DoF) decoupled linear model, which is compatible with the ILC approach. We implemented the data-driven ILC design method on our customized over-actuated UAV platform in both simulation and real-world experiments, which demonstrated the ILC's ability in learning and compensating for repeatable aerodynamic effects, such as downwash effects and fixed wind field disturbances. To the best of our knowledge, this is the first time that repeatable aerodynamic effects have been compensated in a feedforward manner to achieve superior performance.