Disturbance observer-based adaptive neural guidance and control of an aircraft using composite learning

An adaptive neural guidance and control system is proposed in this paper for a generic fixed-wing aerial robot. Unlike most of the existing low-level control systems, which utilize a non-adaptive guidance loop, in this work both the guidance and control loops are trained using an efficient adaptive neural algorithm. A feedforward neural network is employed in each loop to identify uncertain dynamics, while an adaptive disturbance observer allows to compensate for both the external disturbances and estimation error of the neural network. This would lead to a resilient flight control system, and thus, the asymptotic stability of both the guidance and control loops can be theoretically ensured for a generic aerial robot subject to different types of nonparametric internal and external disturbances. Besides, to enhance the learning efficiency, a composite learning method is adopted in which the neural network and the disturbance observer are trained using a composite error function consisting of the tracking error and the estimation error of an introduced adaptive state observer. To the best of the authors’ knowledge, this is the first completely adaptive integrated guidance and control system with guaranteed stability under parametric and nonparametric internal and external disturbances. The introduced control system is then applied to a simulation model of an electric aircraft that has been validated on the basis of real data and flight experiments. The obtained results indicate that the proposed approach could be considered a reliable guidance and control system for a generic fixed-wing aerial vehicle in the presence of actuator faults, unmodeled dynamics, external disturbances, and measurement noises.