Fast Parameter Convergence in Adaptive Flight Control

Increasingly, a number of problems in control of complex systems require fast learning of uncertainties, with only a finite excitation of the system dynamics. Uncertainties may additionally be localized, requiring algorithms which provide for fast uncertainty estimation in specific locations in a state space. During the learning process, reference tracking performance must additionally be maintained for safety-critical systems. In this paper we present numerical simulation examples demonstrating learning of nonlinear in parameter uncertainties which occur due to ground effect present during aircraft landing. We demonstrate the efficacy of a recently developed parameter update algorithm with time-varying learning rates in estimating the unknown ground effect. In particular we show fast uncertainty estimation with a finite excitation, in addition to localized learning and long-term learning properties.