Info

World Congress on Unmanned Systems Engineering

July 2014

Abstract

In this paper, an anti-swing manoeuvre controller for a rotorcraft unmanned aerial system with an attached suspended load (slung-load) is proposed. The presented architecture is powered by Echo State Networks that enables a simple modeling of the controller and outperforms linear techniques in terms of robustness to unmodelled dynamics and disturbances. The external load behaves like a pendulum; this can change the natural frequencies and mode shapes of the low frequency modes of the RUAS. The technique chosen to solve the problem is to achieve both robust performance and computational efficiency. Reservoir computing is an alternative to gradient descent methods for training Recurrent Neural Networks. Echo State Networks is a type of reservoir computing. It has been demonstrated that a model and controller design using echo state networks may be developed. ESN performs well to control unknown nonlinear systems.