Hi,
Selecting the H∞ controller weighting parameters is one the most important part of designing of such a good controller, however, there is no specific method in the most references to do that and most of the works have been done by the trial and error methods.
Do you have any suggestion in order to tune the H∞ controller?
Thanks.
yes, this is one of the important questions in the field of H0infinity control theory. I think you can use GA, PSO, and other swarm optimization algorithms to achieve this mission after minimizing certain performance index. But, the attached thesis is doing auto-tuning for the H-infinity parameters, it is a PHD thesis, it is very important for you to start with.
good luck
take a look on my PhD thesis:
Thesis Polynomial and State-Space Based H-Infinity Robust Controlle...
I used trial-and-error to tune the performance weighting filters...
An approximate procedure is described in ...
M. Grimble, "Robust industrial control systems : optimal design approach for polynomial systems", Hoboken, NJ, Wiley, 2006
Simply tuning a controller so that it has a "healthy" phase margin (e.g. 60 deg) usually ensures that it also has a low H-inf norm, e.g. see ...
Conference Paper Optimal Design of Digital Anticipatory Servomechanisms with ...
I think Trail and error is the good than other methods. The available methods give you values of the parameters of the transfer function of W1(s) and W2(s) but it ultimately may not works and may fail to give you feasible controller. It is again required to adjust by trial and error. Thus I think Trail and error is the best way.
Hi,
Thank you very much for your constructive comments.
@ Ibraheem K. Ibraheem:
What is the best performance index, which can be defined and minimized by the optimization algorithm (cost value for optimization algorithm)?
@ Debasish Mondal :
How do you recognize that you select the best adjusting parameters for your controller by trial and error method? in this way, always there is a doubt whether you select the best adjusting parameters or not.
Thanks.
Right!
After getting controller you then apply to your system to investigate its performance.
Again you pls forget about the term best because H-inf design method is a sub-optimal controller design method. Your best controller may not be best for other.
@ Debasish Mondal :
"Your best controller may not be best for another" that's right, the problem is:
how to tune our controller for a specific control problem in a way that no one can tune that controller better for that specific control problem.
Think about a control benchmark problem that a lot of designers try to tune the same controller for that problem, each one select different adjusting parameter for the same control problem and the same controller by trial and error method. all of them think that they tune their controller as best as possible, however, one of them tune the controller better than others and always there is a doubt whether they select the best adjusting parameters or not for that benchmark control problem. therefore, by trial and error you can not tune your controller as best as possible for a specific control problem except by chance.
Do you disagree?
thanks
I think you are absolutely right. In heuristic methods also you can get best results by chance.
Hi, Hmmmm, Performance index may be the ITAE, ISE,.... of the o/p of the closed-loop system best regards
Dear Mr. Hosseini,
The parameters can be selected by trial and error. Furthermore, you can use the criteria defined in the following paper to see the influence of parameters on your performance. You could consider more criteria according to your control objectives.
Equation (5.47) and correspondingly Figure 5.16 in:
https://core.ac.uk/download/pdf/158619926.pdf
- Badges
- Science topic