KeyError 10 in Constraint line of Python code; MCLP Optimization model building

Answered

Tanmoy Das

November 22, 2022 05:02 Edited

• 
• 
• 

I am trying to implement a constraint of the maximal coverage problem (Equation 16.2 from the image at the bottom of this post). A minimal Python code is below. 

But the code is giving me error KeyError: 10 in Constraint1 (bold line in the code). Explanation of Keyerror seems like certain index values might not exist. I check my data, all index values iterating c,f are available in cartesian_prod. 

model1 = gp.Model('facility_location')
select = model1.addVars(num_customers, vtype=GRB.BINARY, name='Select')
assign = model1.addVars(cartesian_prod, ub=1, vtype=GRB.CONTINUOUS, name='Assign')
constrain1 = model1.addConstrs((assign[c,f] >= select[f] for c,f in cartesian_prod), name='Constraint1')

 

Related to

• Python

0

• 

  Jaromił Najman

  • Gurobi Staff

  November 22, 2022 08:15

  Could you please add to your code, how you define \(\texttt{num_customers}\) and \(\texttt{cartesian_prod}\)?

  1
• 

  Tanmoy Das

  November 22, 2022 11:17

  • 
  • 
  • 

  num_facilities = len(facilities)
  num_customers = len(customers)
  cartesian_prod = list(product(range(num_customers), range(num_facilities)))

  0
• 

  Jaromił Najman

  • Gurobi Staff

  November 22, 2022 11:50

  You are constructing your cartesian product over num_customers x num_facilities. In your constraint you access it via c,f in cartesian_prod, i.e., f is from num_facilities. But you constructed select over num_customers, thus you get a key error. You could try either adjusting your cartesian_prod list or how you construct the constraint.

  cartesian_prod = list(itertools.product(range(num_facilities),range(num_customers)))

  or

  constrain1 = model1.addConstrs((assign[f,c] >= select[f] for f,c in cartesian_prod), name='Constraint1')

  Best regards, 
  Jaromił

  0
• 

  Tanmoy Das

  November 29, 2022 20:02 Edited

  • 
  • 
  • 

  Revised formulation below:

  # MIP model formulation
  model1 = gp.Model('MCLP v1')
  # Decision variables
  cover = model1.addVars(pairings, vtype=GRB.BINARY, name='Cover')
  select = model1.addVars(num_facilities, vtype=GRB.BINARY, name='Select')
  # Objective function
  objective1 = gp.quicksum(cover[c, f]*demand[c] for c,f in pairings)
  model1.setObjective(objective1, GRB.MAXIMIZE)
  # Constraints
  c1 = model1.addConstrs((cover[c,f] <= select[f] for c,f in pairings), name='c1')
  c2 = model1.addConstr((gp.quicksum(select[f] for f in range(num_facilities)) == T_maxStations), name='c2')
  #c3 = model1.addConstrs((cover.sum(c,'*') == 1 for c in range(num_customers)), name='c3')
  model1.write('Outputs/model1.lp')
  model1.optimize()

  I got the following output where each c is covered twice. I need c to be covered only once. Constraint c3 (commented in current code) would to the job, but, c3 makes model1 infeasible. 

  ['cover',
   MultiIndex([(0, 14),
               (0, 15),
               (2, 14),
               (2, 15),
               (3, 14),
               (3, 15),
               (4, 14),
               (4, 15),
               (5, 14),
               (5, 15),
               (6, 14),
               (6, 15),
               (9, 14),
               (9, 15)],
              ),
   'select',
   Int64Index([14, 15], dtype='int64')]

   

  Expected output is something like below:

  ['cover',
   MultiIndex([(0, 14),
               (2, 14),
               (3, 15),
               (4, 14),
               (5, 15),
               (6, 15),
               (9, 14),,
              ),
   'select',
   Int64Index([14, 15], dtype='int64')]

  I think index in Constraint1 in this Python implementation is NOT accurately reflecting the math model.

  0
• 

  Jaromił Najman

  • Gurobi Staff

  November 30, 2022 09:33

  The example you provided is not a minimal reproducible example. There are definitions missing, e.g., pairings, num_facilities, demand.

  Did you analyze the LP file you write to \(\texttt{Output/model1.lp}\)? Maybe you can spot some mistake in there. It might be useful to reduce the size of your model when analyzing the LP file.

  0
• 

  Tanmoy Das

  December 01, 2022 05:04 Edited

  • 
  • 
  • 

  Hello Jaromil,

  I will keep your suggestions in mind. 
  I solved the conflict of C1 & C3 by changing comparison operator == into <= in C3 

  model1.addConstrs((cover.sum(c,'*') <= 1 for c in range(num_customers)), name='c3')

  0

Please sign in to leave a comment.