Re: [Help-glpk] Piecewise linear objective functions

[Andrew Makhorin]

> I would like to know if it is possible with GLPK to solve problems
> with a piecewise linear objective function.

It is possible. However, you need to reformulate your problem,
because glpk does not support non-linear objective and constraints.
A standard way to model a piecewise objective is using SOS2
constraints, which, in turn, can be modeled as follows:
http://lists.gnu.org/archive/html/help-glpk/2007-06/msg00005.html

> I currently work on a model for supply chain planning which uses a
> term in the objective funktion with is equal to zero as long as the
> value of a decision variable is greater than a given constant value,
> and wich is equal to the product of a cost constant and the
> difference of the constant and the decision variable.

> I use ILOG OPL Studio for developing the model and CPLEX for solving
> it.

Please note that glpk is much weaker than cplex in the sense of
ability to solve large or hard MIPs.

> Recently I got the hint to look for GLPK, so I made it running on my
> PC and ported the model. Everything is fine except the piecewise
> objective term.

> (In ILOG OPL I modeled it with a function within the objective
> function:

> maximize sum(...) ... + cost * maxl(0, constant - Variable);  )

> BTW: is there any GUI for GLPK which has similar functionality like
> OPL Studio for CPLEX?

Unfortunately, not.