Iterative Learning Control for Precise Aircraft Trajectory Tracking

Abstract

In this thesis, an iterative learning control method to improve precision in commercial aircraft trajectory tracking is proposed. Given a four-dimensional trajectory to be followed, the proposed method improves the system performance in following the trajectory using the spatial and temporal deviations suffered by previous aircraft to anticipate recurring disturbances, represented by weather conditions and unmodeled system dynamics, and proactively compensate for them, so that the subsequent aircraft intending to fly the same planned trajectory will follow it with greater precision than the previous ones. The iterative learning control algorithm is divided into two steps: estimation of the disturbances and model errors affecting the tracking performance, and update of the control inputs. Two different estimation methods are tested and compared: a Kalman filter, which requires careful tuning and relies on prior knowledge about the dynamics of disturbances and model errors, which are assumed to be linear, and a recursive Gaussian process regression, which estimates and predicts disturbances and model errors at a low computational cost without the linear assumption, and without relying on prior knowledge about their dynamics. As regards to the update step, both direct and indirect iterative learning control approaches are considered. Whereas the former directly updates the control input to be used in the following iteration, the latter generates a new reference signal to be fed into the underlying feedback controller of the aircraft to execute the next iteration and therefore is nonintrusive with respect to the avionics of the aircraft. Both update strategies rely on a nominal dynamic model of the aircraft, in which input and state constraints can be explicitly considered, but require the system dynamics to be repetition-invariant. To overcome this limitation, a multi-aircraft transfer learning strategy is proposed, which enables knowledge about learned trajectories to be transferred among dynamically different aircraft at each iteration. For this purpose, the baseline controller of the aircraft is augmented with a model reference adaptive controller, which forces the aircraft to behave close to a given reference model. The obtained results show a significant reduction of the trajectory tracking error in few iterations, proving the effectiveness of the iterative learning control method applied to commercial aircraft trajectory tracking. The proposed method is suitable to be used in busy terminal maneuvering areas, in which the time-based separation between aircraft is short enough to expect similar weather conditions. Higher precision in trajectory tracking implies an increase in the predictability of aircraft trajectories and an improvement in the efficiency of the air traffic management system. Airlines can also benefit from this higher predictability by reducing the number of alterations when following their designed trajectories, which entails a reduction of costs and emissions.