Is there a hinfsyn replacement?

Is there a replacement/updated version of the function hinfsyn? It looks like hinfsyn is now classed as obsolete(?) - if I run:

which hinfsyn

one of the subdirectories in the file path is called 'rctobsolete'. I ask this as the function doesn't seem to work properly. Depending on the generalised plant I've defined, P, sometimes:

[K,CL,gam,info] = hinfsyn(P,ny,nu);

works as expected, but sometimes it spits out variables K, CL, gam and info that are just empty arrays(!) - it doesn't display any warnings or errors, it just produces nothing. Does anyone know what's happening here?

Is there a newer 'replacement' function for H-infinity synthesis?

Thanks, Oliver

Answer 1

[Janak Thotakura]

The 'hinfsyn' function is good to use, usually the documentation page shows warning when a function is planned to be replaced in the future release.

When the 'hinfsyn' function returns empty, it means that it did not find any controller that could meet your specifications.

I have informed the relevant team to consider adding a warning or throw an error in this case in a future release.

Hope this helps,

Janak

[Oliver Dellar]

Okay, thanks for the info. Yes I just found it a bit strange that it wouldn't throw an error or warning upon failing to compute a controller!

Cheers, Oliver

Answer 2

[Pascal Gahinet]

No it's not going away, in fact it went through a major overall in R2018b, see

https://www.mathworks.com/help/robust/release-notes.html

By default "which" points to the @double implementation, which has been obsolete for many years. Use

which -all hinfsyn

to see all implementations. The @DynamicSystem method, which is called when P is a state-space model object (@ss), is the current version.