Dissertation
Models and methods for mobile facility routing
University of Iowa
Doctor of Philosophy (PhD), University of Iowa
Spring 2026
DOI: 10.25820/etd.008447
Abstract
Mobile facilities offer flexibility and accessibility advantages over fixed infrastructure across diverse service applications. In practice, these operations involve spatiotemporal uncertainty that affects service delivery -- from customer behavior in retail logistics to demand variability in humanitarian aid to infrastructure damage in disaster response. This dissertation develops optimization models and solution methodologies that explicitly model these uncertainties to improve mobile facility routing in three operational settings.
First, reverse logistics for retail returns motivates a scheduling problem where high-resolution daily customer mobility patterns inform mobile facility location over time. Using GPS path data, we develop a multinomial logit choice model with period-specific parameters to maximize the expected number of customer visits, and through a variable neighborhood search approach we demonstrate that path-aware mobile facility scheduling is most valuable in tightly capacity-constrained environments with period-based customer detour sensitivity.
Second, humanitarian aid distribution under stochastic demand introduces the challenge of equitably allocating nonreplenishable capacity across multi-period routes. We formulate this problem by maximizing the net benefit of service subject to service level balance constraints. We solve the formulation with variable neighborhood search integrated with gradient descent to jointly optimize discrete routing and continuous capacity deployment decisions. Computational experiments across 232 instances achieve an average optimality gap of 0.59\% while revealing asymmetric sensitivities to coverage region misspecification.
Third, uncertain network topology compounds the challenges of disaster response operations. We formulate this setting as a Markov decision process in which edge accessibility and site demand are revealed through exploration. We develop a rollout-based approach with information relaxation to address these coupled uncertainties. Additionally, we discover graph attention networks provide a promising value function approximation for this application.
Details
- Title: Subtitle
- Models and methods for mobile facility routing
- Creators
- Jessa Rhea
- Contributors
- Jeffrey W. Ohlmann (Advisor)Beste Basciftci (Committee Member)Samuel Burer (Committee Member)Ann Melissa Campbell (Committee Member)Barrett W. Thomas (Committee Member)
- Resource Type
- Dissertation
- Degree Awarded
- Doctor of Philosophy (PhD), University of Iowa
- Degree in
- Applied Mathematical and Computational Sciences
- Date degree season
- Spring 2026
- DOI
- 10.25820/etd.008447
- Publisher
- University of Iowa
- Number of pages
- ix, 153 pages
- Copyright
- Copyright 2026 Jessa Rhea
- Language
- English
- Date submitted
- 04/19/2026
- Description illustrations
- Illustrations, graphs, charts, tables
- Description bibliographic
- Includes bibliographical references (pages 139-148).
- Public Abstract (ETD)
Unlike fixed facilities, mobile service facilities can relocate to provide service directly to the populations they support. They may operate as mobile customer service units, health clinics, or disaster relief supply vehicles, among other applications. This dissertation develops improved routing methodologies for mobile facilities by recognizing that each application faces various forms of uncertainty in where and when service is needed.
In retail returns, a primary challenge is anticipating where customers are willing to detour. Using smartphone location data that capture natural daily movement patterns, we design schedules that position mobile return centers along common travel paths, increasing convenience while accelerating the recovery of returned products. In humanitarian applications such as mobile food pantries or medical clinics, uncertainty arises from limited supplies and unknown community needs. We develop routing strategies that allocate resources equitably across stops rather than exhausting supplies early, ensuring all visited communities receive service. Disaster response introduces further complexity, as road accessibility and benefactor demand may be unknown until responders arrive. We develop adaptive routing policies that coordinate a fleet of mobile facilities to learn which roads are passable and serve people quickly.
The common thread in this work is that mobile facilities must make routing decisions before knowing exactly what they will encounter. By developing mathematical models accounting for uncertainty in customer behavior, demand for resources, and infrastructure disruption, this research provides a foundation for deploying mobile services in unpredictable environments.
- Academic Unit
- Interdisciplinary Graduate Program in Applied Mathematical & Computational Sciences
- Record Identifier
- 9985177075602771
Metrics
1 Record Views