PhD Thesis Defense: Pushpendra Singh

ZOOM LINK

"Transportation Systems Planning: From Network Design to Operational Routing"

Abstract

Transportation network design and operations constitute a well-studied and practically important area of research, with applications spanning roads, railways, public transit, shared micromobility, and airline operations. Across these settings, planners must decide where infrastructure should be located, how networks should be connected, and how assets should be routed through the resulting system. These problems are challenging because they involve large-scale discrete decisions, operational constraints, and complex interactions between supply and user demand. This thesis develops new optimization models and solution approaches to address these challenges across multiple transportation applications.
 
First, we focus on shared micromobility systems, such as bike-sharing and e-scooter-sharing, which can improve first- and last-mile connectivity, replace short car trips, and enhance the sustainability of urban transportation systems. Their design is challenging because strategic decisions, such as station locations, capacities, and fleet sizes, interact closely with operational decisions, such as rebalancing; moreover, the passenger demand depends on service quality. To address these interdependencies, we develop an integrated mathematical and computational framework that captures strategic design, operational approximations, and passenger choice. We develop a non-convex mixed-integer nonlinear optimization (MINLO) model, and propose a fast spatial-decomposition heuristic that produces near-optimal solutions in substantially shorter runtimes than existing approaches.

Transportation network design problems become especially difficult when passenger choice is embedded in the model, since demand then depends on the attractiveness of the designed network. In the second part of this thesis, we develop a general-purpose non-convex MINLO framework for this class of problems and propose a solution approach based on reweighted L1-norm minimization, which significantly outperforms competing benchmark methods in computational performance while maintaining high solution quality.

The third part of this thesis focuses on the tail assignment problem in airline operations, which consists of assigning specific aircraft tails to scheduled flights while satisfying aircraft continuity and operational restrictions. We study two formulation paradigms, arc-based and string-based, and develop a new parallelized accelerated column generation approach, particularly to handle maintenance-regulated settings.

Overall, we address infrastructure placement, network connectivity design, and operational asset routing, through new models and scalable algorithms, while preserving essential system features and computational tractability.

Thesis Committee

• Vikrant Vaze (Chair)
• Geoffrey Parker
• Wesley Marrero
• Luis Cadarso (Rey Juan Carlos University, Madrid Spain)

Contact

For more information, contact Thayer Registrar at thayer.registrar@dartmouth.edu.