Stochastic Planning and Scheduling with Logic-Based Benders Decomposition

We apply logic-based Benders decomposition (LBBD) to two-stage stochastic planning and scheduling problems in which the second stage is a task. solve master problem with mixed integer/linear programming subproblem constraint programming. As cuts, we use simple no-good cuts as well analytic develop for this application. find that LBBD computationally superior integer L-shaped method. In particular, branch-and-check variant of can be faster by several orders magnitude, allowing significantly larger instances solved. This due primarily computational overhead incurred method while generating classic from continuous relaxation an subproblem. To our knowledge, first application optimization second-stage comparison The results suggest could promising approach other robust or combinatorial recourse. Summary Contribution: study important class problems, namely, programs recourse, are known extremely difficult general. focus on problem, literature best knowledge. Our exemplifies how one exploit structure derive novel them within algorithm. proposed algorithm solves intractable commercial solvers state-of-the-art decomposition-based methods, such believe will inspire further research hybrid methods solving problems.