Abstract—In this investigation, a neural network approach is presented for dynamics identification of a single-bladed aerial vehicle or a monocopter. Implementation of neural networks let us do the non-parametric identification process regardless of the system dynamics. Here, we have initially designed a feedforward network and found that this approach is insufficient for the mentioned purpose. Therefore, a novel network with NARX structure with one hidden layer, tansig activation function and 15 neurons is designed and excellent results are obtained due to consideration of past outputs in the training process.
Index Terms—Monocopter, dynamics identification, feedforward neural network, NARX.
The authors are with the Department of Aerospace Engineering, Sharif University of Technology, Tehran, Iran (e-mail: kebriaee@sharif.ir).
[PDF]
Cite: Mostafa Ezabadi, Mohammad Hasan Sabeti, and Afshin Banazadeh, "Dynamics Identification of a Monocopter Using Neural Networks," International Journal of Modeling and Optimization vol. 7, no. 3, pp. 179-183, 2017.