• Gurobi Staff

Hi Jordi - Did you try to compute an IIS as discussed in the Knowledge Base article How do I determine why my model is infeasible?

yes, but still it is not very clear to me what is wrong the ISS:

\ Model CVRP_copy
\ LP format - for model browsing. Use MPS format to capture full model detail.
Minimize

Subject To
R7: C12 + C13 + C24 + C25 + C36 + C37 + C48 + C49 + C60 + C61 + C72 + C73
= 1
R42: C12 + C24 + C36 + C48 + C60 + C72 = 1
R43: C13 + C25 + C37 + C49 + C61 + C73 = 1
Bounds
Binaries
C12 C13 C24 C25 C36 C37 C48 C49 C60 C61 C72 C73
End
AC

• Gurobi Staff

You probably need to define N and V as

    N = [i for i in range(1,n)] # set of clients    V = [i for i in range(n)] # set of clients plus the depot (depot is 0)
This would lead to a feasible model. But I think a constraint is missing that the same vehicle is used when arriving at client i and when leaving client i.
And finally, I think
# Each vehicle must leave the depot    mdl.addConstrs(quicksum(x[i, 0, k] for i in V if i != 0) == 1 for k in range(num_vehicles))

needs to be

# Each vehicle must leave the depot  mdl.addConstrs(quicksum(x[0, i, k] for i in V if i != 0) == 1 for k in range(num_vehicles))
Then it also fits your evaluation of the solution.