CODEWITHSUNDEEP
mathematical-optimizationcplexampl
Axelitama
Axelitama

Reputation: 33

AMPL CPLEX: is not an indicator constraint message

I have this AMPL model

model;

param n >= 1 integer;

set V = 1..n;
set T = 1..n;
set R = {(i, j) in {V, V}: i <> j};

param capacity >= 0;

var m{V} binary;
var path{T, R} binary;
var u{T} binary;

minimize obj:
    sum{t in T} u[t]
;

subject to max_capacity{t in T}:
    u[t] = 1 ==>
        sum{(i, j) in R} m[j] * path[t, i, j] <= capacity
;

but when I try to solve it using cplex it returns this message:

CPLEX 20.1.0.0: logical constraint _slogcon[1] is not an indicator constraint.

What does this message mean? Is there a simple way to fix the problem?

Upvotes: 0

Views: 363

Answers (2)

Alex Fleischer
Alex Fleischer

Reputation: 10059

In AMPL like in OPL you could use CP.

using CP;

int n=10;

range V = 1..n;
range T = 1..n;

tuple t
{
  int i;
  int j;
}
{t} R={<i,j> | i,j in V:i!=j};

int capacity=4;

dvar boolean m[V];
dvar boolean path[T,R];
dvar boolean u[T];

minimize 
    sum(t in T) u[t]
;

subject to 
{
  forall(t in T)

    (u[t] == 1) =>
        (sum(<i,j>  in R) m[j] * path[t, <i, j>] <= capacity);
      }        
;

works fine and relies on CPOptimizer

If you prefer to use MIP you can rewrite your logical constraint with bigM

int n=10;

range V = 1..n;
range T = 1..n;

tuple t
{
  int i;
  int j;
}
{t} R={<i,j> | i,j in V:i!=j};

int capacity=4;

float bigM=100;

dvar boolean m[V];
dvar boolean path[T,R];
dvar boolean u[T];

minimize 
    sum(t in T) u[t]
;

subject to 
{
  forall(t in T)

    //(u[t] == 1) =>
        (sum(<i,j>  in R) m[j] * path[t, <i, j>] <= capacity*u[t]+bigM*(1-u[t]));
      }        
;

NB:

I used OPL because I love OPL but you can do the same with AMPL

Upvotes: 1

Erwin Kalvelagen
Erwin Kalvelagen

Reputation: 16714

That is not a very intuitive error message, to put it mildly. Developers underestimate how important informative error messages are. A good message would have prevented this post.

I suspect this is because the constraint is not linear (but rather quadratic). Cplex (like other solvers) only allows linear indicator constraints.

Upvotes: 1

Related Questions