Research

Debojjal Bagchi

My work centers on three interconnected domains: trucking infrastructure, port–hinterland systems, and traffic assignment modeling. Within these areas, I use operations research techniques and data-driven methods to study and improve multimodal freight and logistics networks. The selected publications below highlight projects that best reflect my research interests and the balance I aim to strike between methodological contributions and real-world empirical applications.

Beyond publications, I am a strong advocate of open source research and strive to release large-scale frameworks along with documentation for community use. Scroll down to get a list of those. This page also includes my conference presentations and journal publications, which I update regularly with links to supplementary material and code for replicating results (though they may not always be fully up to date). For an up-to-date list, you can download my resume.

Research Domains

I am interested in how ports connect to their hinterlands and how bottlenecks in one component ripple across the system. My focus includes capacity measurement, anchorage and terminal queues, resilience during disruptions, re-routing, and the interaction of waterside, terminal, and landside operations. I am also drawn to questions of data quality and integration, as well as economic and policy aspects such as port pricing, demurrage, environmental costs, and the safety and sustainability of large-scale multimodal freight systems.

Some publications that reflect these interests include:
  • •
    Measuring capacities in multimodal maritime port systems with anchorage queues Bagchi, D., Bathgate, K., Mitchell, K. N., Asborno, M. I., Kress, M. M., and Boyles, S. D. (2025)
    Port capacity models often fail to distinguish between sustainable long-term operations and absolute maximum throughput during demand surges. We develop a framework to estimate both: operating capacity using a parsimonious queueing-theoretic model, and ultimate capacity by fitting differential equations to simulation outputs. Applying this to the Port of Houston, we show how to identify critical bottlenecks across different conditions, finding that liquid-bulk terminals limit stable operations while pilot availability dictates capacity during major disruptions.
    [View Paper]
I am interested in routing and scheduling problems in freight logistics, especially extensions of classic vehicle routing problems (VRPs). These often involve multiple objectives and practical constraints such as time windows. My interests also include modern contexts like last-mile deliveries and electric freight systems.

Some publications that reflect these interests include:
  • •
    We study energy-efficient and safe routing for last-mile delivery using electric freight vehicles. We account for regenerative braking and explicitly model left turns at intersections, which is tricky since each turn depends on three connected nodes. We frame the problem as a bi-criterion Steiner Traveling Salesperson Problem with time windows, develop exact mixed-integer programming models and local-search heuristics, and validate our methods on benchmark instances and Amazon delivery data from Austin.
    [View Paper] [View Code]
I am interested in the efficiency and robustness of traffic assignment models, particularly under imperfect or uncertain data. This includes probabilistic route choice models like stochastic user equilibrium, methods for quantifying error when algorithms terminate early, and approaches to managing queue spillback in dynamic traffic assignment. More broadly, my interest lies in building data-efficient, reliable models that can inform real-world planning.

Some publications that reflect these interests include:
  • •
    Most traffic assignment algorithms stop using rule-of-thumb measures like relative gap, but these do not guarantee how close the solution is to equilibrium. We derive upper bounds for stochastic user equilibrium that show how far any feasible solution can be from equilibrium in terms of flows and costs. These bounds give a rigorous way to assess convergence and can serve directly as stopping rules in traffic assignment. We show that the bounds are tight, converge linearly near equilibrium, and drastically can cut runtimes in network design applications.
    [View Paper] [View Code]
  • •
    Localized Queue Spillback with Uncertain Demand Robbennolt, J., Bagchi, D., and Boyles, S. D. (2025)
    Dynamic traffic assignment models that include spillback are realistic but depend heavily on accurate demand forecasts, while those without spillback are robust but overlook critical congestion. We develop a localized spillback method that applies spillback only on links with a high likelihood of congestion, identified from demand scenarios. On the Austin network, this approach produces more reliable results than either full spillback or no spillback, while keeping the computational cost reasonable.

Publications & Presentations

Publications

Papers under review

Working papers

Conference presentations