Jorge Mortes Alcaraz thesis defense

Abstract : This thesis seeks to enhance the realism of transportation problems by incorporating urban infrastructure constraints and approaching these problems with data-driven algorithms. Specifically, we focus on two contemporary challenges: the last-mile delivery problem and a demand-responsive public bus system. For the first one, we present the multiple time window learning problem. To extract and learn, we employ multi-armed bandit algorithms. The results indicate that implementing a learning mechanism leads to more realistic and effective routing solutions. For the second one, we address the static on-demand bus routing problem. We develop a heuristic approach based on a small and large neighborhood search metaheuristic combined with a set covering component. The proposed algorithm outperforms existing methods on a new set of instances based on real-world data from New York City. Secondly, we address the dynamic on-demand bus routing problem. We solve it employing one-step look-ahead algorithms, which anticipate potential future events to enhance routing decisions. The proposed policies are evaluated on New York City instances, demonstrating that anticipatory strategies significantly improve decision making by leveraging expected future information.