Adding constraint in gurobipy-pandas with arrays of different lengths

Answered

Koen Timmermans

May 05, 2023 13:37 Edited

• 
• 
• 

Hi,

Using \(\texttt{gurobipy-pandas}\), I want to create the following constraint:

\[ z_{k} \geq y_{ik} \quad \forall i \in A, k \in K \]

I have created y like this (ik_values_y as a multi index):

y = gppd.add_vars(model, ik_values_y, vtype=GRB.BINARY, name="y")

>>>
i   k
5   0     <gurobi.Var y[5,0]>
    1     <gurobi.Var y[5,1]>
    2     <gurobi.Var y[5,2]>
    3     <gurobi.Var y[5,3]>
    4     <gurobi.Var y[5,4]>
                 ...
25  0    <gurobi.Var y[25,0]>
    1    <gurobi.Var y[25,1]>
    2    <gurobi.Var y[25,2]>
    3    <gurobi.Var y[25,3]>
    4    <gurobi.Var y[25,4]>
Name: y, Length: 105, dtype: object

And z looks like:

z = gppd.add_vars(model, k_values, vtype=GRB.BINARY, name="z")

>>>
k
0    <gurobi.Var z[0]>
1    <gurobi.Var z[1]>
2    <gurobi.Var z[2]>
3    <gurobi.Var z[3]>
4    <gurobi.Var z[4]>
Name: z, dtype: object

Currently, I'm trying to add the constraints with

gppd.add_constrs(model, z.groupby("k"), ">=", y.groupby(["i", "k"]))

This returns the following error message: \(\texttt{All arrays must be of the same length}\).

In the previous version of my code (without gurobipy-pandas), I added the constraint like this:

for i in A:
            for k in K:
                model.addLConstr(z[k] >= y[i,k], name=f"G1-{i},{k}")

Hopefully someone can help me out here. Thank you in advance!

Best regards,

Koen Timmermans

Related to

• gurobipy-pandas

0

• 

  Mario Klostermeier

  May 05, 2023 20:40

  • 
  • 
  • 

  Hello,

  this is not perfect but you don't need two loops.

  for i in A:
      model.addLConstr(z >= y[i,:], name=f"G1-{k}")

  0
• 

  Riley Clement

  • Gurobi Staff

  May 08, 2023 00:12

  Hi Koen,

  We need the indices on the series to align to use gurobipy-pandas.  One option is to use pandas.DataFrame.merge to get the alignment although this will drop the "i" index from the resulting dataframe which is not ideal.  There are workarounds to keep the "i" index but it gets clunky.

  A better approach is to align the index of z with y using pandas.Series.reindex

  G1_cons = gppd.add_constrs(  
      model, z.reindex(y.index, level=0), GRB.GREATER_EQUAL, y, name="G1-",
  )

  - Riley

  0

Please sign in to leave a comment.