Decision Optimization

Originally posted by: NguyenHuuTRi

Hello,

I am a beginner of CPLEX optimizer. Actually, I have to revise my model and switch it to CP in stead of MIP. Below is my model. I appreciate it if you could help me how to revise my model in CP way.

Thanks.

using CP;

range type = 1..14;
range course = 1..518; 
range room = 1..150;
range lecture = 1..353;
range timeslot = 1..12;
range group = 1..252;
int lecturetime[lecture][timeslot] = ...;
int courselecture[course][lecture]=...;
int caparoom[type]= ...;
int coursegroup[course][group]=...;

tuple course_type{
    int idcourse;
    int idtype;
};
setof (course_type) coursetypes = ...;

dvar boolean assCourseTimes[course][timeslot];
dvar boolean assGroupTimes[group][timeslot];
dvar int compactGroup[group][1..6];

maximize sum (c in course, t in timeslot) assCourseTimes[c][t];
minimize sum(g in group, t in 1..6)compactGroup[g][t];
 
 subject to {
 

forall (t in timeslot,c in course )  { 
     assign_CourseinTimeslot: assCourseTimes[c][t] <= sum(l in lecture) (lecturetime[l][t]*courselecture[c][l]);
}

forall (c in course){
    OneCourseinTime: sum(t in timeslot)assCourseTimes[c][t] <= 1;
}

forall (t in timeslot, ty in type){
    capa_room: sum(<c,ty> in coursetypes) assCourseTimes[c][t] <= caparoom[ty];
}

forall (t in timeslot, g in group){
    assign_GroupTimes: sum(c in course) assCourseTimes[c][t]*coursegroup[c][g] == assGroupTimes[g][t];
}

forall (t in timeslot: t<7, g in group){
  abs(assGroupTimes[g][t] - assGroupTimes[g][t+6])* 2  <= 
compactGroup[g][t];

forall (t in timeslot: t>6, g in group){

  abs(assGroupTimes[g][t] - 1) <= compactGroup[g][t-6];
};

Hi,

you should use staticLex in order to do multi objective.

You should replace

maximize sum (c in course, t in timeslot) assCourseTimes[c][t];
minimize sum(g in group, t in 1..6)compactGroup[g][t];

by

dexpr float objMax =sum (c in course, t in timeslot) assCourseTimes[c][t];
dexpr float objMin = sum(g in group, t in 1..6)compactGroup[g][t];
maximize staticLex(objMax, -objMin); 

regards

Originally posted by: NguyenHuuTRi

Hi Alex,

Thanks for your answer.

I have another question regarding constraints for CP as I can use the above contraints for CP problem as I have to use other syntax for CP and change all of them.

Thanks.

Tri.

Hi,

the constraints you wrote work both for CP and MIP. But you can improve some by using some specific CP constraints.

For example, you wrote

forall (c in course){
    OneCourseinTime: sum(t in timeslot)assCourseTimes[c][t] <= 1;
}

which works both for CP and MIP, but for CP something like

forall (c in course){
    OneCourseinTime: count(all(t in timeslot)assCourseTimes[c][t],1) <= 1;
}

could work better.

Regards

Originally posted by: NguyenHuuTRi

Thanks Alex.

Anyway, I wonder if I have to set the search space for CP as I have known that there is required to set that to improve the results and time consumption. If so, could you guide me how to do so in my current model.

Thanks so much.

Tri.

Originally posted by: GGR

Hi

You have to declare the execution limit, the number of thread or the number of branches to stop the search in case of non proven optimal incumbent. Create a parameter file in your OPL project and fill the required field (generally a time limit). The IHM is self explanatory.

Hope that helps