15h30 - 15h55
Benders decomposition for tree-of-hubs location problems
In this talk, we study the Tree-of-Hub Location Problems. First, we present relevant literature review and motivation for the problem. We present a mathematical formulation and a Benders Decomposition to solve the problem. In the second part of the talk we study the hop-constrained THLP. We reformulate the problem and present a Benders decomposition to solve the hop-constrained version of the problem. We discuss optimality-feasibility cuts and their relation to other valid inequalities for both variations of THLP. We present experimental results assessing the performance of the proposed solution methodologies. Finally, we draw conclusions and talk about possible future research.
15h55 - 16h20
Binary search for partitioning graphs using k-connected subgraphs
We study the problem of partitioning a weighted graph G = (V, E, w) into p subgraphs, each of which must be k-connected. The weight of a cluster is the minimum weight of a k-connected subgraph, and the objective is to find the partition that minimizes the maximum such weight. We solve this problem using a binary search algorithm, for which the subproblems are solved by means of branch-and-price. Preliminary results will be provided
16h20 - 16h45
A new branch-price-and-cut algorithm for the pickup and delivery problem with time windows
The pickup and delivery problem with time windows aims at finding routes to satisfy a set of requests. We investigate the impact of disregarding precedence and paring relations within the routes to obtain less restrictive dominance rules in column generation algorithms. Multi-commodity paths are then generated to reinforce route feasibility.
16h45 - 17h10
Solving the optimum communication spanning tree problem
This talk presents an exact algorithm based on Benders decomposition to solve the optimum communication spanning tree problem. It integrates a strong reformulation, combinatorial bounds, in-tree heuristics, fast separation algorithms, and a tailored branching rule. Computational experiments show solution time savings of up to two orders of magnitude compared to state-of-the-art exact algorithms.