Proactive Vehicle Routing to Solve the Dynamic Bike Sharing Rebalancing Problem with Potential Extension to Shared Station Vehicle Systems.

Dissertation by Robert Regue
Adviser: Professor Will Recker


Shared-use mobility systems, which enable users to have short-term access to
transportation modes on an on-demand basis, have experienced tremendous growth over
the last decade. However, most of the existing systems suffer from two co-founding
issues: the lack of modeling tools to understand, simulate and predict their behavior and
the lack of integration with the existing transit network. To address those issues, this
dissertation focuses on investigating the operational challenges of bikesharing systems,
with an emphasis on the rebalancing operations and the modeling of a new mobility
concept, Car2work, which builds upon existing carsharing ideas and successfully
integrates with existing transit networks. A methodological framework to solve the
bikesharing rebalancing problem is proposed. The novelties of the approach are that it is
proactive instead of reactive, as the bike redistribution occurs before inefficiencies are
observed, and uses the outputs of a demand-forecasting technique to decompose the
inventory and the routing problem. The decomposition makes the problem scalable,
responsive to operator inputs, and able to accommodate user-specific models. Simulation
results based on data from the Hubway bikesharing system show that system performance
improvements of 7% in the afternoon peak could be achieved.
Car2work main goal is to connect commuters with workplaces while leveraging
the line-haul capabilities of existing public transit systems and guaranteeing a trip back
home, efficiently tackling the “last mile” problem that is a limiting characteristic of
public transit. It differs from the traditional dynamic-ridesharing approaches because it is
designed for recurrent commuting trips where commuters announce their (multiple) trips
in advanced and an automated all-or-nothing matching strategy is performed,
guaranteeing a ride home. The problem is formulated as a pure binary problem that is
solved using an aggregation/disaggregation algorithm that renders optimal solutions. The
solution approach is based on decomposing the problem into a master problem and a subproblem,
reducing the number of decision variables and constraints. As a result, various
instances of the problem can be solved in reasonable amount of time, even when
considering the transit network. The model could be used to simulate a large-scale
implementation of the concept.