Implementing Feasibility Cuts in Benders Decomposition

Answered

Miguel Peinado-Guerrero

• March 23, 2022 22:00

I am attempting to implement feasibility cuts using with gurobi/C# within the canonical benders decomposition framework. I am solving the primal of the master and the primal of the sub. As such, when formulating optimality cuts, I can use GRB.attr.pi to attain the required information from the dual of the sub. In the case infeasibility in the primal of the sub problem, I need to get the direction of unboundness for the sub's dual and use this information in the feasibility cut. Is this what GRB.attr.FarkasDual provides? Do I need to do any other processing on it?

0

• 

  Jaromił Najman

  Gurobi Staff

  • March 24, 2022 08:35

  FarkasDual and FarkasProof provide a certificate of infeasibility and not the direction of unboundness. To get an unbounded ray, you have to query the UnbdRay attribute, which is only available if parameter InfUnbdInfo has been set.

  0

Please sign in to leave a comment.