I advance optimization theory and logistics. Discover the methodologies, impact, and software driving my research.
The crux of many real-world challenges: decisions across time and in the face of uncertainty.
Sequential decision problems span business, engineering, the sciences, and health care. Researchers and practitioners alike struggle to identify effective policies, let alone optimal strategies.
My work closes this gap. I engineer solutions for dynamic and stochastic optimization problems to expand theory and transform practice.
The curse of dimensionality: as problem size scales, the computation required for optimal decisions explodes. Practical problems quickly outpace these methods.
When sequential decision problems resist conventional solution procedures, my work gauges policies. Performance guarantees set a floor on policy quality. Dual bounds impose a ceiling. A tight gap between the two pins down the unknown optimal value.
I integrate tools across disciplines: dynamic programming, optimal control, Markov decision processes, and reinforcement learning. For example, see "Bounded Backward Induction for Max-Min Dynamic Programming" and "A Rollout Algorithm Framework for Heuristic Solutions to Finite-Horizon Stochastic Dynamic Programs."
Transportation and logistics fuel the global supply chain. These problems trigger computational challenges: combinatorial decision spaces complicate routing, and stochastic demand fluctuates across products, regions, and time.
I combine recent theoretical advances with classical deterministic approaches. For decades, transportation research modeled demand, travel conditions, and customer behavior as fixed. These assumptions simplify analysis but rarely hold in practice. Though deterministic methods often fail in the face of uncertainty, with the right adjustments, they generate strong policies, dual bounds, and insights.
This approach produces problem-specific innovations and general methodological advances. Applications include "Electric Vehicle Routing with Public Charging Stations" and "Dynamic Ridehailing with Electric Vehicles," which leverages developments in deep neural networks.
My research in dynamic and stochastic optimization underpins methods to mitigate a spectrum of societal and industrial challenges. These include:
with Luca Bertazzi. April 2026.
How to hedge against worst-case outcomes when making decisions in sequence. An alternative to adjustable robust optimization for discrete and non-convex problems. The approach derives general dual bounds and policy performance guarantees. Think of it as branch-and-bound for dynamic programs.
Read the Paper
with Maximiliano Cubillos, Ola Jabali, & Elena Tappia. May 2025.
Unobserved arrival time distributions, ETAs, a lookahead policy, information relaxations, and an information penalty. This is a story about connecting stochastic dynamic programming with mathematical programming to reduce lead times at warehouses and across the logistics chain.
Read the Paper
with Nicholas D. Kullman, Nikita Dudorov, Martin Cousineau, & Jorge E. Mendoza. To appear in INFORMS Journal on Computing.
Do you remember your first video game console? We remember ours. Decades ago, they provided hours of entertainment. We repurposed them to solve dynamic and stochastic optimization problems.
Read the Paper
with Marlin W. Ulmer & Barrett W. Thomas. Transportation Science, 2024:58(2), 394-411.
Take back your time. Stop waiting around for the cable guy! The paper prescribes a strategy to shrink service windows in the face of uncertain demand without eroding customer service levels.
Read the Paper
with Steve Scroggins, Enbal Shacham, & Tasnova Afroze. Vaccines, 2023:11(1).
How to allocate a limited supply of vaccines during a pandemic: Sift through terabytes of cell phone data, predict disease spread with regression, then distribute across time and space with off-the-shelf optimization software. Help us cut through the red tape by sharing this with policy makers wherever you live.
Read the Paper
with Nicholas D. Kullman, Martin Cousineau, & Jorge E. Mendoza. Transportation Science, 2022:56(3), 567-798.
Transportation Science Paper of the Year
Autonomous ridesharing transforms public transit. The paper introduces AI methods that match riders with vehicles and proactively position the fleet for future use and charging needs.
Read the Paper
with Nicholas D. Kullman, Aurelien Froger, & Jorge E. Mendoza. INFORMS Journal on Computing, 2021:33(4), 1259-1684.
When and where to charge EVs. These tasks underlie many operational problems in sustainable transportation. The paper describes software to optimize charging along a given route.
Read the Paper
with Nicholas D. Kullman & Jorge E. Mendoza. Transportation Science, 2021:55(3), 637-659.
INFORMS Transportation Science & Logistics Society Best Paper Honorable Mention
Don't let the uncertainty of public charging stall fleet electrification. The paper engineers optimization tools to generate delivery routes in the face of congestion.
Read the Paper
with Marlin W. Ulmer, Dirk C. Mattfeld, & Barrett W. Thomas. EURO Journal on Transportation and Logistics, 2020:9(2), 100008.
Rigorous methods outpaced rigorous models, making rigorous science difficult. Route-based Markov decision processes connect dynamic routing problems with the route-based methods used to solve them.
Read the Paper
with Marlin W. Ulmer, Dirk C. Mattfeld, & Marco Hennig. Transportation Science, 2019:53(1), 185-202.
Featured in the 2021 special issue "A Deeper Look at Transportation Science by Topic Area"
On-demand packages and food delivered to our doorstep at a moment's notice punctuate modern living. The paper engineers reinforcement learning methods that anticipate demand and build routes across space and time.
Read the Paper
with Barrett W. Thomas & Jeffrey W. Ohlmann. European Journal of Operational Research, 2017:258(1), 216-229.
How to make decisions in sequence and in the face of uncertainty with performance guarantees. The paper unifies and generalizes rollout algorithms.
Read the Paper
with Don Hardaway, Roel Harryvan, & Frank Wang. Communications of the Association for Information Systems, 2016:38(6), 145-156.
The holy grail of business education resides at the intersection of academics and industry. The paper recounts our experience working closely with a major consulting firm in the classroom.
Read the Paper
with Barrett W. Thomas & Jeffrey W. Ohlmann. Transportation Science, 2016:50(2), 591-607.
Dynamic vehicle routing in the face of uncertain demand underpins modern logistics. The paper demonstrates the value of preemptive capacity replenishment within a lookahead mechanism.
Read the Paper
with Marlin W. Ulmer, Dirk C. Mattfeld, & Marco Henig. Logistics Management, 2016, 217-227.
Couriers and parcel delivery firms navigate without knowing when and where demand will occur. The paper designs a method to plan routes now with explicit consideration of future demand.
Read the Paper
European Journal of Operational Research, 2015:241(2), 361-369.
What do waste collection, fuel distribution, and grocery transport have in common? They utilize multi-compartment vehicles. The paper constructs a method to manage multi-compartment logistics in the face of uncertain demand.
Read the Paper
Information Systems and Operational Resesarch, 2014:52(1), 1-9.
Even without politics, elections remain messy. The paper recounts methods for statewide management of volunteer resources during the 2012 US presidential election.
Read the Paper
with Jeffrey W. Ohlmann & Barrett W. Thomas. Operations Research, 2013:61(1), 138-154.
Uncertain demand often thwarts logistics planning, but it doesnāt have to. The paper engineers methods to dynamically adjust route plans in response to demand. It uses static routes to make decisions on the fly.
Read the Paper
with Jeffrey W. Ohlmann & Barrett W. Thomas. European Journal of Operational Research, 2012:217(2), 312-323.
Vehicle routing problems come in many flavors, but often share a similar underlying structure. The paper exploits this structure. The methodology is applicable to a range of problems.
Read the Paper
with Wooseung Jang & Marilyn Rantz. The Gerontologist, 2008:48(3), 338-348.
Part 1 of 2. The paper examines practical takeaways from using Bayesian networks to predict and manage quality of care.
Read the Paper
with Wooseung Jang. Health Care Management Science, 2008:11(4), 382-392.
Part 2 of 2. The paper outlines the technical aspects of using Bayesian networks to predict and manage quality of care.
Read the Paper
Nicholas D. Kullman. Python Package Index.
Use 2018 data from Manhattan as the stage for developing AI ride-sharing strategies. The software accompanies the paper titled "Dynamic Ridehailing with Electric Vehicles."
Get the Code
with Nicholas D. Kullman, Aurelien Froger, & Jorge E. Mendoza. Python Package Index.
The fixed route vehicle charging problem underlies many problems in the management of electric vehicles. The software accompanies the paper titled "Electric Vehicle Routing with Public Charging Stations."
Get the Code
with Robert Hansen, Tristan Thiede, Jeffrey W. Ohlmann, & Barrett W. Thomas. Computational Infrastructure for Operations Research (COIN-OR).
The software implements compressed annealing and simulated annealing, both metaheuristics for combinatorial optimization problems. If you have a basic heuristic for your problem, Djinni can make it better.
Get the Code