Originally posted by: Allexandre

Hi,

I built a Dantzig-Wolfe decomposition model and in each iteration I have to update the pi_[j] value. This  value came from j dual variables from j constraints in my Master Problem. I have to do this several times, inside a loop, until I reach the optimum value. But, It looks like the values that I catch from dual are not being set to the subproblem objective function.

Thanks.

You post looks incomplete. In any case, you may want to try to explicitly re-create the objective function instead of updating it:

// Remove the old objective function.
IloObjective oldObjective = cplex.getObjective();
cplex.getModel().remove(oldObjective);
oldObjective.end();
// Setup a new objective function:
IloObjective newObjective = ...;
cplex.getModel().add(newObjective);

If you just change cplex.getObjective().getExpr() then the change will not be propagated to the IloCplex instance that has extracted the objective. You have to at least to cplex.getObjective().setExpr() to notify the IloCplex instance about changes to the objective function. See also the reference documentation of class IloObjective.

Originally posted by: Allexandre

Hi Daniel,

Here is the link to the image which shows my objective function from Subproblem: https://www.dropbox.com/s/feloazsgw0ciq7m/dantzig-wolfe_sp.jpg

To solve this, I used the IloObjective::setLinearCoef

when I was creating the model, I kept its coeficients equal 1, for x variable. During the running, I set the new coeficients as (c[i,j] - pi[j]) for each variable x.

But now, in the Master Problem, which is minimize an error variable e, which has coefficient equal 1, I have to update the coeficients for the master's constraints. Is there any function for this?

Thanks.

If you store your constraints as an array of IloRange then you can use IloRange::setLinearCoef() or IloRange::setLinearCoefs() to change the coefficient of a variable in a constraint.