Issue while modelling indicator constraints

Answered

Vamsi Krishna

• November 30, 2023 22:04

I am trying to model the following constraint.

where x_ij is binary Decision variable, c_j is continuous decision variable, and k_i is parameter. K is set for the parameter k_i values K ={k_1,......, k_n}

# Indicator constraints for c[j] = K[i] implies z[i, j] = 1

for i in range(len(K)):

    for j in range(m):

        model.addConstr(c[j] - M[i] == 0 >> z[i, j] == 1, name='indi_constr_1')

        # Indicator constraints for c[j] != M[i] implies z[i, j] = 0

        model.addConstr(c[j] - M[i] != 0 >> z[i, j] == 0, name='indi_constr_2')

# Link z variables to x variables

for i in range(len(K)):

    for j in range(m):

        model.addConstr(z[i, j] == x[i, j], name="indi_constr_3")

 

For these constraints, I am getting the following error.

line 48, in <module>
    model.addConstr(c[j] - M[i] == 0 >> z[i, j] == 1, name=f"indicator_constraint_cj_mi_{i}_{j}")
TypeError: unsupported operand type(s) for >>: 'int' and 'Var'

Any help would be much appreciated.

0

• 

  Marika Karbstein

  Gurobi Staff

  • December 01, 2023 07:51

  There are several issues in your formulation

  1. The expressions in the indicator constraints (see Model.addGenConstrIndicator() ) need brackets, e.g., 
     

     model.addConstr((x7 == 1) >> (x1 + 2*x2 + x4 == 1.0))

  2. The indicator variable needs to be binary
  3. not-equal constraints are not supported

  So, you need to reformulate your expression. For example, c != k could be expressed as |c-k| >= eps.
  Note, to use Model.addGenConstrAbs() you will need helper variables.
  The following articles might be helpful

  • How do I model logical and general expressions?
  • Why doesn't Gurobi support strict inequality constraints like x < a?
  • How do I model conditional statements in Gurobi?

  0

Please sign in to leave a comment.