Originally posted by: SystemAdmin

[Didier Vidal said:]

MJEber,

Maybe you should add a term in your objective that measures how "evenly product are distributed over time instead ".
This way the model would make the difference between two solutions with similar costs, but a better distribution over time.
The difficulty will be to pick the right weight for this term.

Didier.
#DecisionOptimization
#MathematicalProgramming-General

Originally posted by: SystemAdmin

[oussedik said:]

Hi,

You may consider ILOG consulting services who can support you here in implementing the constraints and objective in such a way that products are evenly distributed. Achieving a good equilibrium between resolution performance and an even distribution of products can also be important in such cases. Moroever, other aspects of the models can also be raised and discussed with the consultants.

Sofiane
#DecisionOptimization
#MathematicalProgramming-General

Originally posted by: SystemAdmin

[MJEber said:]

[quote author=Didier Vidal link=topic=643.msg1965#msg1965 date=1225270972]
MJEber,

Maybe you should add a term in your objective that measures how "evenly product are distributed over time instead ".
This way the model would make the difference between two solutions with similar costs, but a better distribution over time.
The difficulty will be to pick the right weight for this term.

Didier.


Numerical analysis showed that the ships were all lining up and waiting for the most profitable customer.  So I totalled WaitTime and added that to my objective which helped spread the distribution a little.  I now have only one out of five customers untouched.

How would you suggest I implement distribution over time?  That is what I'd like to do but I'm really creating a headache trying to figure out the expression and objective statement.
#DecisionOptimization
#MathematicalProgramming-General

Originally posted by: SystemAdmin

[dgravot@noos.fr said:]

To model your fair distribution , would it be possible to minimize the distance to the median (or the average) between your variables ?

Something like


range horizon=1..100;
int average = 1;
dvar float x[horizon];
dvar float m;

dexpr float+ u[horizon];
dexpr float+ v[horizon];

minimize sum(t in horizon) u[t]+v[t];

constraints
{
forall(t in horizon)
x[t] == m + u[t] - v[t];

if(average==1) //otherwise, m is the median
sum(t in horizon) u[t] - v[t] == 0;
}


Hope this helps
#DecisionOptimization
#MathematicalProgramming-General