If I have designed a fuzzy inference system (say controller) using Fuzzy Logic Toolbox in Simulink. Now Based on some performance criteria, I need to update/modify the existing structure of fuzzy controller, say, Rule base, scaling factors and membership functions.
How do I do so in Simulink? Lets say, I have implemented the algorithm in a S-function/Embeded Mat;ab Function block in Simulink. now, I want to access the rule base and MFs and update them online without stopping the simulation.
As per my understanding, you can try following,
1. Use user function block, in that update FIS file.
2. Write fuzzy controller code in .m file and use user function block.
Please try above stuff, let me know, if any issue.
In matlab command : write fuzzy, then from edit option go import from file.
Then select the particular file which you have to edit or change then edit it.
@Bikas Mondal
Thanks for your answer. But I thats not what I want. I want online tuning of fuzzy controller in Simulink during running a single simulation task. What you are saying that I know but that is manual edit which can be done in Toolbox itself.
@Anirudh Nath
actually in FIC, there are few rulebase. As per rulebase FIC works.If you change or modify the controlled signal then you have to edit or modify rulebase. So there are no such tuning option for FIC.
@Bikas Mondal
You didn't get my question actually. What you are saying, is ok. I am working with fuzzy PID controller and ANFIS and I have the knowledge to work with that.
Actually, I want some heuristic techniques like GA,PSO etc. to optimize my fuzzy controller [membership functions shape, Rule base]. So, I want to implement this in a simulink environment. How to do that in Simulink?
Hi Anirudh,
The easiest way to learn about using Fuzzy Logic Toolbox in Simulink is to read the user's guide in Matlab which tells you everything you want to do in fuzzy logic. I attach the doc of Matlab user's guide for you herewith.
Also, thought you might like to have a look at one of my papers in RG using the rule base, scaling factors and membership functions of a .fis file in Simulink just as what you want. https://www.researchgate.net/publication/270103983_Fuzzy_based_direct_torque_and_flux_control_of_induction_motor_drives
Sorry, the attached file is too big to be uploaded into RG. So, I attach the Mathworks web link instead. Actually, Fuzzy Logic Toolbox in Simulink is pretty easy to do. Hope this helps.
Good luck - Ning
http://au.mathworks.com/help/releases/R2015a/pdf_doc/fuzzy/fuzzy.pdf
Conference Paper Fuzzy based direct torque and flux control of induction motor drives
Hey,
As far as I know, there is no way to do this exclusively in Simulink. You will have to use some code.
For this, you will generally have to become familiar with manipulating fuzzy structures in an .m file, with functions such as 'writefis', 'savefis', etc. (see link):
http://www.mathworks.com/help/fuzzy/mamdani-fuzzy-inference-systems.html
Then , there are two options:
1.) If your simulation is simple enough, you could probably get away with just coding it and running it as a function. The fuzzy controller you would evaluate with 'evalfis' and the rest of the system with 'tf' for its transfer function. The problem here is that you cannot define the fuzzy controller as a transfer function (that's part of the beauty of fuzzy), so you can't really get an overall deterministic transfer function for the system. Also, the controller needs to work with errors calculated using data from the previous time step. So if you want to run the simulation over a period of time, this might not work.
2.) What you will probably end up doing is running a Simulink simulation from your script using the 'sim' command. This lets you run your chosen model in Simulink and returns your target data (just replace your final Sink with a 'To Workspace' block). Your evolutionary algorithm can then run the simulation each iteration, get the results and test against those for fitness. I refer you to the following thesis I found online:
ftp://ftp.ucauca.edu.co/Facultades/.FIET_serepiteencuentasyocupaespacio/DEIC.back/docs/Materias/Control%20Inteligente/proyectos/GAs/brutonj/JByrne_Meng_03.pdf
(JP Byrne, GA-Optimization of a Fuzzy Logic Controller)
As regards the actual optimization algorithm in Matlab, you may or may not be able to use their standard optimization functions (such as 'ga', 'pso', etc.), probably depending on how much of the problem you can cram into the fitness function that you pass to the algorithm. In the end, it might be worth your while to simply code the thing yourself, thus giving you total control over the implementation.
Also, remember to check out the documentation ('help ') for all the functions I've mentioned, as there is obviously a whole lot of details you will need to understand when you use them.
Hope this helps. Hit me back with any questions or to let me know how it goes.
Victor
@Victor Vladareanu
Thank you very much for your kindness. I will let you know after I try to implement it.
@Anirudh Nath
Im working with Fuzzy Logic and i need something like you. Were you able to modify the rules (tunig) ? if Yes. Could you explain me how? Regards
- Badges
- Science topic