AUVSI’s Unmanned Systems.
Controller design for aircraft with unusual configurations presents unique challenges, particularly in extracting valid mathematical models of the MRUAVs behaviour. System Identification is a collection of techniques for extracting an accurate mathematical model of a dynamic system from experimental input-output data. This can entail parameter identification only (known as grey-box modelling) or more generally full parameter/structural identification of the nonlinear mapping (known as black-box). In this paper we propose a new method for black-box identification of the non-linear dynamic model of a small MRUAV using Echo State Networks (ESN), a novel approach to train Recurrent Neural Networks (RNN).