Multi-Aircraft Transfer Learning for Aircraft Trajectory Tracking in Continuous Climb and Descent Operations

This paper presents a control technique for aircraft trajectory tracking that combines Iterative Learning Control (ILC) with Model Reference Adaptive Control (MRAC). ILC enhances the accuracy of aircraft in following a predefined trajectory based on the deviations in space and time observed in previous flights to anticipate repetitive disturbances proactively. However, ILC requires the system to be repetition-invariant, which is not applicable in real operations where different aircraft consecutively perform flights along the same trajectory. To address this drawback, a multi-aircraft transfer learning strategy is considered. At each iteration of the ILC, this strategy allows learned trajectory knowledge to be transferred among different aircraft. The proposed approach involves augmenting the baseline aircraft trajectory tracking controller with an MRAC, which ensures that the aircraft behaves similarly to a given reference model, while the ILC acts as a high-level adaptation scheme, compensating for repetitive disturbances affecting the flight. Numerical experiments are conducted using various simulated aircraft following the same trajectory in continuous climb and descent operations. Results show that the MRAC-ILC method outperforms the combination of ILC with the baseline feedback controller without MRAC augmentation, achieving a substantial reduction in the trajectory tracking error after a few iterations. This improvement remains consistent even in the presence of model uncertainties, disturbances, and changing aircraft dynamics. In summary, the MRAC-ILC method makes ILC suitable for real operations, enhances the predictability of aircraft trajectories, and consequently improves the efficiency of the air traffic management system.