currently I am trying to implement an simple branch and cut algorithm using the GAMS Branch & Cut Facility and CPLEX. In a simple implementation a relaxed problem formulation is strengthened in branching process by providing some additional cuts (Problem VRP).
A relaxed MIP problem formulation is given and a subtour identification routine has been defined. The call of the cut routine works out as 2 additional cuts are added to the problem (Log file: *** Calling cut generator. Added 2 cuts). In contrast to the log file output, somehow these provided cuts are not getting effective as:
- solving goes on ignoring the new cuts added directly providing the global solution and
- changing values of sense_c does not have any effect on the solution process as well
No further branching / solution evaluation takes place. After stating the "proven optimal solution" following content (fourtimes, for each variable) is printed to the log file: Variable x_ij(Customer1,Customer2) does not exist in model which (most probably) seems to be an error but is not highlighted as one. Of course I have checked syntax, variables and parameters again and again and again to resolve this. Finally, a comparison of bchin.gdx and bchout.gdx shows no differences to me. Also numcuts, sense_c, rhs_c and x_ij_c are looking properly for me. It seems like the introduction of the created cuts into the model is not working as expected. Would be happy for any advice/ hint what might go wrong here.
I am using version 30.1.0 and have attached a sample implementation with sample data reproducing the problem.