New exact and heuristic methods for a collection of logistics problems

This thesis addresses logistical challenges across energy, routing, and education through novel decision models and solution techniques. It emphasises fast, effective decision-making in environments with incomplete information to support strategic and operational objectives.

The first theme focuses on network design problems under demand uncertainty, subject to a service level constraint. A novel FlowMIS formulation enhances Benders' decomposition to solve network design problems with service level constraints more efficiently. Numerical experiments show significant speed-ups and the ability to handle larger problem instances.

The second theme considers vehicle routing problems. We explain the development of PyVRP, a hybrid genetic search metaheuristic that achieves state-of-the-art performance while remaining user-friendly. Its application in urban waste collection demonstrates significant potential cost reductions through integrated container selection and vehicle routing decisions.

The final theme addresses personalised learning in secondary education, focusing on hourly activity scheduling. An adaptive large neighbourhood search metaheuristic, compared against optimal solutions in a large study, achieves near-optimal results. We further evaluate structural and staffing changes in education related to personalised learning.