Originally posted by: SystemAdmin

[golopang said:]

Dear All,

May I know if there is any simple way that I can solve a MILP formulation (in LP text format) by using the CP engines??? Actually, I tried using CPLEX to solve the MILP formulation, but the problem is too complex for CPLEX to find even a feasible solution, so I am thinking if CP can help to identify a feasible solution faster. Thanks.

Regards,
Anthony PANG
#ConstraintProgramming-General
#DecisionOptimization

Originally posted by: SystemAdmin

[jfk said:]

Hello,
Constraint programming has a sort of special branch called global solvers or interval solvers where they deal with continuous variables. Since you posted here your question I assume when you ask CP you mean ILOG CP. As far as I know ILOG CP doesn't do interval solver.

on the other hand I'm very interested to know this problem of yours which can't be solved by CPLEX. would you share it? thanks in advance...

cheers
#ConstraintProgramming-General
#DecisionOptimization

Originally posted by: SystemAdmin

[golopang said:]

Dear JFK,

Thanks for your quick response. Actually I am doing a research on Ship routing and berth assignment. And I have the MILP formulation with about 3400 integer decision variables for a sample problem size. By using the CPLEX, it takes me 24 hours but still no first solution is found. Any idea??

What I am actually concerning is whether there is any simple way to read a MILP formulation (which I have done it for CPLEX) but solve by Constraint Programming engine. Thanks.

Anthony

#ConstraintProgramming-General
#DecisionOptimization

Originally posted by: SystemAdmin

[Didier Vidal said:]

Anthony,

If your problem contains only discrete variables (no continuous variable), you can run it with CP Optimizer as is.
But in general, you need to change the formulation to get the best of CP Optimizer.
For MILP, modelers often use boolean variables. In CP, modelers tend to use integer variables with the full domain instead.

Here is an example:

Classical assignment MIP

var int xm[L,R] in 0..1;

forall (k in L) sum (r in R) x[k,r] = 1

forall (r in R)
sum (k in L) x[k,r] = 1


Classical assignment CP

var R xc[L];



alldifferent(xc);



Assign many to few MIP

var int xm[L,R] in 0..1;

forall (k in L) sum (r in R) x[k,r] = 1

forall (r in R)
sum (k in L) x[k,r] = Rsizes[r]



Assign many to few CP

var R xc[L];


forall(r in R)
  count(xc, r) == Rsizes[r];



Hope this helps,

Didier.
#ConstraintProgramming-General
#DecisionOptimization