Optimization of on-demand shared mobility: passenger and freight applications

On-demand mobility and logistics platforms have transformed urban transportation over the past decade by enabling app-based, door-to-door passenger and parcel services. However, their rapid growth of dedicated vehicle fleets has intensified congestion and environmental externalities in already capacity-constrained cities. Optimizing these services and integrating them with existing road users presents open challenges, often leading to computationally demanding problems. This thesis addresses these challenges through two complementary optimization frameworks that enhance capacity utilization in passenger and freight systems.

The first part studies dynamic ride-pooling in large-scale ride-sourcing platforms. Standard batch matching approaches are inherently myopic and offer limited ability to anticipate spatial supplyâ demand imbalances, particularly during demand surges when idle vehicles are scarce. To overcome this limitation, this thesis proposes a two-level optimization framework that indirectly rebalances in-service vehicles through strategic matching decisions. An upper-level maximum flow model anticipates inter-regional demandâ supply evolution, while a lower-level Integer Quadratic Programming formulation performs batch matching with soft adherence to regional flow targets. This implicit rebalancing mechanism improves system resilience and supports flexible pooling strategies without relying on explicit idle vehicle relocation.

The second part introduces a multimodal urban parcel delivery framework that integrates drones with scheduled bus services. We formulate the Pick-up-and-Delivery Problem with Drones and Scheduled Lines (PDP-DSL), in which drones serve first- and last-mile legs while parcels may use public transport for middle-mile transfers. The problem is modeled as an arc-based mixed-integer linear program. To address computational complexity and overcome well-known limitations to commercial solvers, this work develops tailored metaheuristics and feasibility checks based on Bellmanâ Ford and dynamic programming.

Together, these contributions demonstrate how hierarchical optimization, spatial aggregation, mathematical modelling, and metaheuristics can address operational constraints and challenges in large-scale mobility systems, emphasizing the complementarity of simulation and optimization for problems of different scales. By improving the utilization of existing transportation capacity, the proposed approaches provide methodological foundations for sustainable passenger and freight integration in increasingly multimodal urban environments.