What are the parameters you want to estimated? And how they relate the inputs to output?

If you answer these questions, then it is possible to say which system identification method is suitable for it.

You should be careful as you have closed-loop control, only some classes of system identification methods are applicable.

hi Yassine el Houm ,

you can read my article, I used the same method i.e. n4sid . I hope it helps you. However, generally, I have to say that it is very important to choose the order in a rational way. here it is the link to my article, you can freely download it in RG:

Article Experimental Parametric Identification of a Flexible Beam Us...

I think you should work in the linear region. I mean that the roll and pitch should be in a range of +/- 20 degree. In this case you will avoid the sine and cosine nonlinearities.

Hope you all the best

Hi Yassine. I'm assuming the 3 degrees of freedom of the test bench is pitch, roll, yaw. If that assumption is true then a black box approach used for the estimation of the nonlinear dynamics cannot be predicted accurately given lack of excitation of the coupled dynamics.

You can refer to " Introduction to Multicopter Design and Control " Chapter 12

Hi Paulin Kantue

I tried to excite the system using a pseudo random binary sequence , still the same issue though ;(

the input command is :

U[k]=Upid[k]+Uprbs[k]

Yassine el Houm, then I'll quantify how much Uprbs[k] gets rejected by your controller to establish control loop 'stiffness'. In closed loop system identification, a badly designed controller can prevent any system identification method from being successful.

If you can get your hands on: Flight Vehicle System Identification: A Time-Domain Methodology, Second Edition by Ravindra V. Jategaonkar.

There's quite a few principles in there (without too mach maths overload) that could be helpful.