Disturbance-Learning-Based Robust Model Predictive Control for Attitude Tracking of Small Aircraft

Abstract

Attitude control of small aircraft under unknown disturbances poses a tricky task. This article proposes a model predictive controller (MPC) for small aircraft based on online disturbance learning to enhance attitude tracking accuracy. A known nominal model is used to predict the system’s behavior. Adaptive radial basis function (RBF) neural networks, employing an improved gradient descent with momentum, are recommended for learning unmodeled dynamics and disturbances. Subsequently, a MPC integrating disturbance-learning-based Lyapunov constraints is devised. Control constraints are realized through an auxiliary control unit, and its design process relies on the Lyapunov comparison principle. The controller’s recursive feasibility and practical stability are proven. The experiments were conducted using the small aircraft platform, validating the controller’s efficacy in this article.