Abstract: Designing high quality two-tier transportation systems requires sophisticated modeling techniques and solution strategies. This thesis focuses on designing three such transportation systems with a two-tier network structure.

The first project presents a service region design methodology for first-and-last-mile on-demand services connecting passengers' origins or destinations with mainline public transit stops. The total system-wide cost for both passengers and transit system providers is explicitly considered. This work provides a new approach to effectively and efficiently allocate shuttle vehicles to passenger demands in a region with one or more mainline stops serving as first-and-last-mile depots. The proposed approach achieves a considerably lower total cost objective while also using a computationally more tractable algorithm.

The second project contributes to planning a truck-and-drone collaborative parcel delivery scheme. It integrates a comprehensive trajectory modeling procedure into a traveling salesperson problem with drone (TSP-D) to minimize total system-wide travel time cost. It utilizes a multiobjective nonlinear optimization model and leverages machine learning based solution techniques, to develop an original systematic computational framework. This framework applies a neural network model to approximate physics-based drone trajectories, in turn outperforming all state-of-the-art benchmarks that either oversimplify drone physics properties or encounter intractability in high-dimensional combinatorial solution generations.

The third project incorporates the concept of optional stops in the traditional public transit system routes. It optimizes the selection of optional stops from a pool of candidate optional stops, to form a flexible-route public transit system. This project explores the benefits and tradeoffs suggested by a flexible-route structure through an optional stop set selection mechanism in the hope of combating the emerging transit crisis. With a two-stage stochastic programming model and solution algorithms sequentially optimizing optional stop selection and operational routing decisions, this work is able to increase public transportation ridership and accessibility in real world cases at practical instance sizes without compromising the reliability of the bus transit systems.

Thesis Committee: Prof. Vikrant Vaze (Chair), Prof. Eugene Santos Jr., Prof. Geoffrey Parker, Prof. Armin Fügenschuh (external)