Sparse online Gaussian process adaptation for incremental backstepping flight control

Abstract

Presence of uncertainties caused by unforeseen malfunctions in actuation or measurement systems or changes in aircraft behaviour could lead to aircraft loss-of-control during flight. This paper considers sparse online Gaussian Processes (GP) adaptive augmentation for Incremental Backstepping (IBKS) flight control. IBKS uses angular accelerations and control deflections to reduce the dependency on the aircraft model. However, it requires knowledge of the relationship between inner and outer loops and control effectiveness. Proposed indirect adaptation significantly reduces model dependency. Global uniform ultimate boundness is proved for the resultant GP adaptive IBKS. Conducted research shows that if the input-affine property is violated, e.g., in severe conditions with a combination of multiple failures, the IBKS can lose stability. Meanwhile, the proposed sparse GP-based estimator provides fast online identification and the resultant controller demonstrates improved stability and tracking performance.