Defense of Afaf ALOULLAL

Distribution systems are essential for moving goods and supporting economic growth, requiring designs that are efficient, resilient and adaptable to growing and uncertain demands.

This PhD thesis focuses on the optimization of distribution network structures, specifically addressing the problems of hub localization and goods routing (HLRP). HLRP takes up the challenge of structuring transport and distribution flows between multiple sources and destinations.

The strategic choice of hub locations enables flows to be consolidated and redistributed, resulting in cost savings and improved delivery times.

While much research has focused primarily on hubs location problems, recent interest in HLRP highlights its importance in practical applications.

This thesis examines a single allocation model with capacity constraints for hubs and a general network topology.

The first phase of this research introduces time-dependent decision making for HLRP by dividing the planning horizon into several periods and developing a four-phase matheuristic method. This approach, combining relax-and-fix, variable neighborhood descent and local branching techniques, shows significant cost savings when integrating multi-period parameters.

The second phase develops a fast algorithm operating independently of commercial solvers to handle large instances. A variable neighborhood search metaheuristic (GVNS) is proposed for single-period HLRP, incorporating ten neighborhood structures and an adaptive penalty mechanism. Different configurations of neighborhood orders, perturbations and penalties have been tested to select the best performing variants, demonstrating the effectiveness of GVNS in large instances.

The final phase addresses flow uncertainty in HLRP using a chance constraint model.

An approximate solution algorithm, based on Monte Carlo simulation and sampled average approximation (SAA), has been proposed and tested, integrating the methods developed in the previous research phases.

Comparative analysis with deterministic chance constraint models confirmed the combined power of simulation and metaheuristics to produce high-quality stochastic solutions.

This thesis establishes a basis for integrating multi-period aspects and stochastic flows into the hubs localization and goods routing problem, offering valuable perspectives for future research.

• Mr Abdelhakim ARTIBA, Université Polytechnique Hauts de France, Thesis co-director
• Mrs Hande YAMAN, Faculty of Economics and Business, KU Leuven, Rapporteur
• Mr Stefan NICKEL, Karlsruhe Institute of Technology (KIT), Rapporteur
• Mr Justo PUERTO, Universitat de Sevilla, Examiner
• Mrs Hatice CALIK, KU Leuven, ELECTA & EnergyVille, Examiner
• Mr Olivier PETON, IMT Atlantique, LS2N laboratory (UMR CNRS 6004), Examiner
• Mr Raca TODOSIJEVIC, LAMIH UMR CNRS 8201 - DPT AUTOMATIQUE, Thesis co-supervisor
.
• Mr Francisco SALDANHA DA GAMA, Sheffield University Management School, Thesis co-supervisor
  .