Section: New Results

Aggregation algorithms for network flow mathematical models

Participant : F. Clautiaux.

We designed a general framework for solving very large network flow mathematical models that use a pseudo-polynomial number of variables. It is based on an initial aggregation of the vertices of the model and its iterative refinement using different optimization techniques. This led to large improvements for a special case of vehicle routing problem. Several theoretical questions regarding convergence, worst-case analysis and approximation algorithms are raised by our work and are now under study.