Small-scale helicopter system identification model using recurrent neural networks

Designing a reliable flight control for an autonomous helicopter requires a high performance dynamics model. This paper studies the recurrent neural network nonlinear model identification of a small scale helicopter. We have selected a Nonlinear AutoRegressive with eXogenous Inputs SeriesParallel (NARXSP) network model which identifies the dynamics model of an unmanned aerial helicopter from real flight data. The identification process is conducted by using the well known Levenberg-Marquardt learning algorithm. The obtained dynamics model shows good fitness with the actual data. This accuracy might be used to realize a reliable flight control for an autonomous helicopter.