GA toolbox error
in [Matlab]
|
Prev: MathWorks Webinar: Tips and Tricks: Data Analysis and Surface Fitting with MATLAB"- 5/18/10
Next: How can I change the precision in a figure
From: Arvind Kumar on 14 May 2010 13:11 I m trying to desgin a controller which has 5 parameters using the GA toolbox. My fitness function needs to run a simulink model and get its output to obtain the error from the reference. Because GA uses a wide range of populations, the simulink model inside the fitness function does not run for some of those values giving the following error "Derivative input 1 of 'ref_ctrl_new/Reformer/Subsystem2/Integrator' at time 10.02145886657267 is Inf or NaN. Stopping simulation. There may be a singularity in the solution. If not, try reducing the step size (either by reducing the fixed step size or by tightening the error tolerances)." This stops the toolbox execution. Is there a way to chose the populations such that it does not produce a singularity in the solution? (OR) Is it possible for the GA to avoid values which produce singularity in the solution? Any help is greatly appreciated.......... Thanks and Regards Arvind Madhavapeddi
From: Alan Weiss on 18 May 2010 12:47
On 5/14/2010 1:11 PM, Arvind Kumar wrote: > I m trying to desgin a controller which has 5 parameters using the GA > toolbox. My fitness function needs to run a simulink model and get its > output to obtain the error from the reference. Because GA uses a wide > range of populations, the simulink model inside the fitness function > does not run for some of those values giving the following error > "Derivative input 1 of 'ref_ctrl_new/Reformer/Subsystem2/Integrator' at > time 10.02145886657267 is Inf or NaN. Stopping simulation. There may be > a singularity in the solution. If not, try reducing the step size > (either by reducing the fixed step size or by tightening the error > tolerances)." > > This stops the toolbox execution. > > Is there a way to chose the populations such that it does not produce a > singularity in the solution? (OR) Is it possible for the GA to avoid > values which produce singularity in the solution? > > Any help is greatly appreciated.......... > > Thanks and Regards > Arvind Madhavapeddi the patternsearch solver is robust to this type of error. I suggest you try it instead of ga. Additionally, you can give bounds on your 5 parameters so that ga or patternsearch will not step into areas that are nonsensical for your problem. Alan Weiss MATLAB mathematical toolbox documentation |