No abstract

This paper presents the results of an online identification algorithm based on Autoregressive models aided by Artificial Neural Networks for the non-linear dynamics of an Unmanned Aerial Vehicle (UAV) platform. Numerical simulations were performed for different combinations of the network structures and the autoregressive model. The weights were trained and updated online using the Levenberg Marquardt method. The results have been validated using the real-time Hardware in the Loop simulation technique for different sets of flight data.

<p class="MsoNormal" style="text-align: left; margin: 0cm 0cm 0pt; layout-grid-mode: char;" align="left"><span class="text"><span style="font-family: ";Arial";,";sans-serif";; font-size: 9pt;">In this paper, real-time system identification of an unmanned aerial vehicle (UAV) based on multiple neural networks is presented. The UAV is a multi-input multi-output (MIMO) nonlinear system. Models for such MIMO system are expected to be adaptive to dynamic behaviour and robust to environmental variations. This task of accurate modelling has been achieved with a multi-network architecture. The multi-network with dynamic selection technique allows a combination of online and offline neural network models to be used in the architecture where the most suitable outputs are selected based on a given criterion. The neural network models are based on the autoregressive technique. The online network uses a novel training scheme with memory retention. Flight test validation results for online and offline models are presented. The multi-network dynamic selection technique has been validated on real-time hardware in the loop (HIL) simulation and the results show the superiority in performance compared to the individual models.</span></span><span style="font-family: ";Arial";,";sans-serif";; font-size: 9pt;"></span></p>

No abstract