Using simulink as fitness function for GA

Hello, I built a PI fuzzy control system using simulink with 3 gains that need to be tuned. I want to tune these gains automatically using genetic algorithm based on minimizing the mean square error. I have a m-file that runs the GA and calls a fitness function in another m-file. The fitness function m-file receives the gains as suggetsed population from GA and it is supposed to run the simulink block after sending the gains to it. The problem is that how to send the population (gains) during the run of GA to simulink and run the simulink block every generation???

FYI I used the sim command to run simulink and to workspace block to get the system output to the m-file. I tried using set_param,assignin commands and it didn't work (I knew because the output gains I got are not optimal for the controller)...

Please help me it's for my master thesis and I need this to finish...

[ashish]

dear hala richard crozier i hv same problem if u read it kindly me also [e-mail]

[Sakshi Bhonsle]

Kindly mail me if your solution for the above mentioned solution was found : [e-mail]

Answer 1

[Richard Crozier]

I think this might help you:

http://stackoverflow.com/questions/5140013/executing-simulink-model-by-m-file-and-retrieving-the-results

Answer 2

[Muruganandham Subramanian]

Hi,

FYI,Genetic algorithm is a population based method, probably the gains for PID controller(i.e, Proportional gain, Integral gain, Derivative gain) are to be splitted into range b/w mininum to maximum (for e.g If you know the PID gains range approximately).For this you can use 'unifrnd' command.then by looping method, for every data of PID gains(i.e depends on population), give as i/p to .mdl, then by GA algorithm, you can choose best results

Answer 3

[Sangeetha Elumalai]

I also want to tune PID controller in simulink by using GA... if anyone knows, pls help me

[Hassan Yousef]

me also

[Zhengyan Mao]

Anyone of you figures this out?

Answer 4

[Stephan Kurzawe]

Hej, I ran into the same issue today. On a matlab example for lsq-nonlin I found the solution:

You just have to add 'SrcWorkspace','Current' into your sim command. So that it looks:

sim('model.slx','SrcWorkspace','Current');

Hope it helps. Stephan