Title: Resilient Inventory Management In Large-Scale Supply Chain Networks

 

Date: Monday, November 30, 2026

Time: 9:00 AM to 11:00 AM ET

Location: George Tower 602 (Physical Internet Center), and virtual via Microsoft Teams

Meeting link: https://teams.microsoft.com/meet/285621886779885?p=m3pb13Efq26UZlu6wM (Meeting ID: 285 621 886 779 885, Passcode: qe9aL6Hx)

 

Ashwin S. Pothen

Operations Research PhD candidate

H. Milton Stewart School of Industrial and Systems Engineering

Georgia Institute of Technology

 

Committee

  1. Dr. Benoit Montreuil (advisor) - Industrial and Systems Engineering, Georgia Institute of Technology
  2. Dr. Matthieu Lauras (co-advisor) - Adjunct Professor, Industrial and Systems Engineering, Georgia Institute of Technology; Professor, KEDGE Business School, Bordeaux, France
  3. Dr. Frederick Benaben (co-advisor) - Adjunct Professor, Industrial and Systems Engineering, Georgia Institute of Technology; Professor, IMT Mines Albi, France
  4. Dr. Eunhye Song - Industrial and Systems Engineering, Georgia Institute of Technology
  5. Dr. Jan C. Fransoo - Tilburg School of Economics and Management, Tilburg University, Netherlands

 

Summary

Modern supply chains operate in complex, disruption-prone environments, involving multiple independent decision-makers and customers with varying willingness to wait and substitute. This dissertation develops methods for resilient inventory management centered on robust autonomy: the ability of facilities, networks, and organizations to keep serving demand on their own during disruptions, while coordinating with others when it helps. Our work draws on large-scale industrial data, operational experience, and educational deployments.

 

Chapter 1 develops a multi-agent, quasi-autonomous decision-support architecture for resilient demand-supply alignment. Using perception and decision agents, we model the operations of a North American manufacturer, integrating demand estimation from transaction records, dynamic forecasting, disruption sensing, and rolling-horizon mixed-integer optimization of production and replenishment. In a retrospective simulation of the manufacturer's network through the 2020 COVID-19 shutdowns, disruption-aware farsighted planning significantly improves profitability and key dimensions of resilience, including resistance, robustness, and recovery, relative to myopic lean approaches.

 

Chapter 2 examines the resilience of inventory policies for partially substitutable products with lost sales, using data from a large North American distributor's product network. We propose the Robust Autonomy Period (RAP) as an intuitive, unifying measure that represents the number of days a node or (sub)network can serve customer demand at r%-robustness. The RAP policy uses daily review to maintain autonomy at the group level, with substitution-aware optimization of individual product mix as demand and lead times evolve. We then examine service, profitability, and resilience through controlled comparisons with well-studied policies during normal operations and supply disruptions.

 

Chapter 3 addresses the management of societally critical products, grounded in our living lab initiative for the supply and distribution of personal protective equipment (PPE) on the GeorgiaTech campus, serving over 200 laboratories during the COVID-19 pandemic. Using simulations, we quantify the impact of technologies that enable quasi-autonomous distribution. The chapter extends to inter-organizational coordination through federated autonomy, where organizations help each other in times of distress by sharing surplus inventory while protecting their local autonomy, as a stepping stone toward full Physical Internet realization. In a representative simulation of a 67-facility network across Georgia serving approximately 1.45 million users, the federated approach significantly improves demand fulfillment during a regional supplier shutdown.

 

Chapter 4 makes a pedagogical contribution through a data-driven educational gaming simulator, deployed with master's students in Supply Chain Engineering at ISyE over three fall semesters (2024 to 2026). The simulator enables students to experiment with inventory and capacity policies and interpret the trade-offs between service and profit. From 13,500 completed runs across three cohorts, statistical analyses show consistent associations between repeated experimentation and improved performance, and each doubling of trial runs is associated with substantial performance percentile gains within cohort.

 

Overall, the dissertation contributes to the design, optimization, operation, and experiential education of resilient inventory management, toward the Physical Internet paradigm of open and hyperconnected logistics and deployment networks.