Methodologies

The main disadvantage of the integration of the problems and of providing a realistic approach is the computational burden and the ability to solve adequately the problem at hand. This will be tackled by using innovative solution methodologies that will be able to provide profitable, sustainable and useful solutions within reasonable time.

As a big picture, optimization models and techniques will be applied, namely decomposition procedures and matheuristics. Matheuristics combine mathematical models - which capture and seize the structure and full requirements of the problem - and metaheuristics - algorithms that systematically explore solution spaces, providing good solutions in reduced time. More specifically, we will be applying co-evolutionary genetic algorithms, which provide adequate and flexible conceptual structures for the integration with decomposition approaches and have shown potential to tackle highly complex problems.

Other methods will be used to deal with specific details within this big picture. To deal with uncertainty, stochastic models will be used and scenarios will be generated to apply in these models. Selecting the best solution in an uncertain environment often depends on the expected value. Other options include selecting the best performance on the worst-case scenario or intermediate solutions, which provide robustness to the solutions obtained. These options will be studied in this project, task that will be facilitated by the structure of co-evolutionary genetic algorithms.

Finally, simulation methods will be used to test the validity of solutions provided and to perform sensitivity analysis to critical elements (e.g. the impact of variations on expected length of vehicle repositioning).