[Coin-ipopt] DynOpt Dependencies

[Andreas Waechter]

The mailing list hasn't been working properly for a while.  I sent this
message a few times earlier, but it never made it through....  Maybe we
are having better luck this time...

---------- Forwarded message ----------
Date: Fri, 19 Dec 2003 11:17:33 -0500 (EST)
From: Andreas Waechter <andreasw at bigopt.watson.ibm.com>
To: calgeri at chem.polimi.it
Cc: coin-ipopt at www-124.southbury.usf.ibm.com,
     Yi-Dong Lang <lang at andrew.cmu.edu>
Subject: Re: [Coin-ipopt] DynOpt Dependencies

Hi Carlo,

Many thanks for your message.

> Hi, I'm an italian academic researcher interested in using DynOpt in order to
> develop a siumlation software (in fortran language) for dynamic optimization of
> a polymerization plant. I would like to know if exist a more detailed
> documentation (or othter examples codes).
> A second important question is related to HSL routines needed: it is true that
> routines MA48 (that is not freely downloadable from HSL) are necessary in order
> to compile dynopt?

Let me answer the second (easier) question first:  Yes, the current
implementation requires MA48, but this is only a matter of what linear
solver call is implemented, and this could be changed with not too much
effort to using MA28.

As for the support of DynOpt, there is no more detailed information, and
as a matter of fact I intended to remove the DYNOPT part for the next
release of IPOPT.  When we originally made IPOPT open source, we included
all components (not only the optimizer but also the dynamic optimization
part), since we were not sure what we wanted to do with it, and also we
didn't know how much interest there was, particularly for DYNOPT (and
there wasn't much so far).

If you are really interested in DYNOPC and it turns out that the full
space approach is not suitable for you, we can try to work something out,
but due to my own time constraints I will not be able to invest much time,
so that my involvement could only be in guiding you in using and/or
adapting the DYNOPT code.  DYNOPT is also part of DYNOPC, which is a
Windwos-based interface to DYNOPT and has been mainly developed by Yi-Dong
Lang (one of the core developer) - maybe he can help you by providing you
with information on DYNOPC?  (Yi-Dong, what is your take on this?)

On the other hand, you might also want to have a look at Tobias
Jockenhoevel's 'OptControlCentre' (see http://www.optcontrolcentre.com/ )
which is a nice tool for modeling and solving optimal control problems,
and which uses IPOPT as the optimizatien engine (in the full space mode).

I'm sorry that I cannot give you a more encouraging answer at this point -
maybe someone else reading this mailing list has an idea or suggestion to
add.

Best regards,

Andreas