Enhancing a branch-and-bound algorithm for two-stage stochastic integer network design-based models

In this paper we present branch-and-bound (B&B) strategies for two-stage stochastic integer network design-based models with integrality constraints in the first-stage variables. These strategies are used within L-shaped decomposition-based B&B framework. We propose a valid inequality in order to improve B&B performance. We use this inequality to implement a multirooted B&B tree. A selective use of optimality cuts is explored in the B&B approach and we also propose a subgradient-based technique for branching on 0–1 feasible solutions. Finally, we present computational results for a fixed-charge network design problem with random demands.