• The complete list of publications

• Google Scholar
• ORCID
• ResearcherID
• DBLP

Selected Publication

Publications

Articles under Review

1. Thi Thao Vy Bui, Yantong Dou, Changhyun Kwon, Yoonjin Yoon, Yu Zhang. Repurposing Existing Helipads for Scheduled Urban Air Mobility: An Optimization Framework for Vertipad Selection and Timetable Design abstract

   Objectives: Urban Air Mobility (UAM) can leverage existing helipads to accelerate deployment while reducing infrastructure investment. This study develops an integrated optimization framework for repurposing existing helipads as vertiports and designing scheduled UAM service timetables.
   

   Methods: A two-stage optimization framework was developed to address the complexity of UAM system planning. In the first stage, a Multi-Objective Integer Linear Programming (MOILP) model with the ϵ-constraint method is used to select optimal vertiports from existing helipads by balancing population coverage as a proxy for demand, community noise exposure, and ground safety risks. In the second stage, fleet timetabling is optimized over a time-expanded network subject to battery dynamics, vertiport capacity, and operational constraints. To solve large-scale scheduling problems, a Hybrid Genetic Algorithm (HGA) was developed, featuring an indirect chromosome representation and an integrated exact Linear Assignment Problem (LAP) decoder.
   

   Findings: Results demonstrate that repurposing existing helipads can support an efficient scheduled UAM network in Seoul. Pareto analysis reveals a clear trade-off between accessibility and safety: maximizing demand coverage substantially increases ground risk, whereas a modest reduction in coverage yields significant safety improvements. The proposed HGA further enables scalable timetable optimization for large UAM networks while satisfying battery and vertiport capacity constraints.
   

   Novelty: The proposed framework integrates strategic vertiport selection with tactical fleet timetabling by considering passenger accessibility, ground safety, community noise, battery dynamics, and vertiport capacity constraints. Additionally, the HGA combines an indirect chromosome representation with an exact LAP decoder, enabling computationally efficient optimization of large-scale UAM scheduling problems while preserving operational feasibility.
   

   Practical Applications: The proposed framework provides a decision-support tool for urban planners, transportation agencies, and UAM operators. By identifying suitable existing helipads for conversion into vertiports and generating operationally feasible service timetables, the framework supports the deployment of scheduled UAM services that maximize accessibility while satisfying safety and environmental requirements.

2. Jaehwan Lee, Changhyun Kwon. Quantum End-to-End Learning for Contextual Combinatorial Optimization arXiv abstract

   Contextual combinatorial optimization (CCO) plays a critical role in decision-making under uncertainty, yet remains a significant challenge. We present Quantum End-to-End Learning (QEL), the first quantum computing-based end-to-end learning framework for CCO that leverages Quantum Approximate Optimization Algorithms. Inspired by the integration of state preparation and evolution in data re-uploading, we propose a context re-uploading phase-separator that jointly captures the complex relations among contexts, uncertain coefficients, and optimal solutions. This allows a contextual encoder to be seamlessly integrated within a quantum surrogate policy, enabling joint end-to-end training with a stationarity guarantee. Exploiting an optimization-aware structure grounded in physical principles that classical methods cannot readily leverage, our approach demonstrates practicality by directly training on task loss despite the discreteness and nonconvexity, while avoiding calls to NP-hard optimization solvers. QEL empirically achieves competitive performance while requiring substantially fewer parameters than classical benchmarks, highlighting its industrial-level potential for the future quantum era.

3. Waquar Kaleem, Doyoung Lee, Changhyun Kwon, Anirudh Subramanyam. Neural Embedded Mixed-Integer Optimization for Location-Routing Problems arXiv abstract

   We present a novel framework that combines machine learning with mixed-integer optimization to solve the Capacitated Location-Routing Problem (CLRP). The CLRP is a classical yet NP-hard problem that integrates strategic facility location with operational vehicle routing decisions, aiming to simultaneously minimize both fixed and variable costs. The proposed method first trains a permutationally invariant neural network that approximates the vehicle routing cost for serving any arbitrary subset of customers from any candidate facility. The trained neural network is then embedded as a surrogate within a mixed-integer optimization problem for location-allocation decisions, which is solved using off-the-shelf solvers. The framework is simple, scalable, and requires no routing-specific knowledge or parameter tuning. Computational experiments on four benchmark sets confirm the effectiveness of our approach. The method achieves median gaps ranging from 1.21% to 7.06% compared to best-known solutions while providing computational speedups of 2× to 120× over state-of-the-art heuristics with particularly strong performance on large-scale instances with up to 600 customers. The findings suggest that the neural-embedded framework is a viable method for tackling integrated location and routing problems at scale. Our code and data are publicly available.

4. Merve Doganbas, Junhak Lee, Hayong Shin, Changhyun Kwon. A Learning-Based Separation Algorithm for the Strengthened Comb Inequalities abstract

   Strengthened Comb Inequalities (SCIs) are a powerful class of cutting planes for the Capacitated Vehicle Routing Problem (CVRP), but their NP-hard separation problem has historically relied on heuristics. This paper introduces CombFormer, the first learning-based framework to generate SCIs. We model the construction of a comb’s handle as a sequential decision problem and train an autoregressive graph transformer-based policy with reinforcement learning to directly maximize the violation of the resulting inequalities. Extensive offline experiments show that CombFormer generates SCIs with significantly higher violations than the widely used CVRPSEP heuristic. The model, trained only on small instances with no more than 100 nodes, demonstrates strong generalization to problems with up to 1000 nodes. When integrated into a branch-and-cut framework, we find that CombFormer’s cuts are complementary to those of CVRPSEP; a hybrid strategy that combines both separators provides the most robust improvements to the root-node dual bound in an iteration-limited setting. CombFormer demonstrates that reinforcement learning is a viable paradigm for discovering high-quality cuts for complex combinatorial valid inequalities.

5. Jae Hyeok Lee, Taekang Hwang, Changhyun Kwon. Asymptotic Bounds for the Traveling Salesman Problem with Drone arXiv abstract

   The asymptotic behavior of the optimal TSP tour length is well known from the classical Beardwood–Halton–Hammersley theorem. We extend this result to the Traveling Salesman Problem with Drone (TSPD), a cooperative routing problem in which a truck and a drone jointly serve customers. Using a subadditive Euclidean functional framework, we establish the existence of an almost sure limit for the optimal TSPD makespan scaled by the square root of the problem size. We derive explicit upper and lower bounds for the speed-scaled Euclidean TSPD model: upper bounds are obtained via structured ring-based tour constructions and Monte Carlo evaluation, while lower bounds are derived from a parametric approach and known bounds on the Euclidean TSP constant. Computational results illustrate how tight the bounds are. We also derive and discuss lower bounds for the Rectilinear–Euclidean mixed TSPD model, in which truck travel is measured by the rectilinear distance and drone travel by the Euclidean distance.

6. Yoonju Sim, Hyeonah Kim, Changhyun Kwon. Test-Time Search in Neural Graph Coarsening Procedures for the Capacitated Vehicle Routing Problem arXiv github abstract

   The identification of valid inequalities, such as the rounded capacity inequalities (RCIs), is a key component of cutting plane methods for the Capacitated Vehicle Routing Problem (CVRP). While a deep learning-based separation method can learn to find high-quality cuts, our analysis reveals that the model produces fewer cuts than expected because it is insufficiently sensitive to generate a diverse set of generated subsets. This paper proposes an alternative: enhancing the performance of a trained model at inference time through a new test-time search with stochasticity. First, we introduce stochastic edge selection into the graph coarsening procedure, replacing the previously proposed greedy approach. Second, we propose the Graph Coarsening History-based Partitioning (GraphCHiP) algorithm, which leverages coarsening history to identify not only RCIs but also, for the first time, the Framed capacity inequalities (FCIs). Experiments on randomly generated CVRP instances demonstrate the effectiveness of our approach in reducing the dual gap compared to the existing neural separation method. Additionally, our method discovers effective FCIs on a specific instance, despite the challenging nature of identifying such cuts.

7. Suyeon Choi, Changhyun Kwon, Seungki Min. Dynamic Pricing of an Expiring Item under Strategic Buyers with Stochastic Arrival arXiv abstract

   We study the optimal dynamic pricing of an expiring ticket or voucher, sold by a time-sensitive seller to strategic buyers who arrive stochastically with private values. The expiring nature creates a conflict: the seller’s urgency to sell before expiration drives price reductions, which in turn incentivize buyers to wait. We seek the seller’s optimal pricing policy that resolves this tension. The main analytical challenge is that buyer type is two-dimensional (valuation and arrival time), which makes equilibrium intractable under general strategies. To address this, we introduce the Value-Based Threshold (VBT) strategy, a tractable framework that decouples these two dimensions. Using this framework, we prove equilibrium existence via an ordinary differential equation and provide a constructive procedure for its characterization. We then derive near-optimal pricing policies for two stylized regimes: a constant price in thin markets and a linear discount in thick markets. Numerical frontier analysis confirms these benchmarks and shows how optimal policy adapts as the seller’s time sensitivity changes. Our findings clarify the conflict between quick sales and strategic waiting. Sellers facing thick markets or high time sensitivity benefit from linear discounts, while in thin markets a constant price neutralizes buyers’ incentive to wait. We also show this simple policy remains robust across broad conditions. For patient sellers, a quasi-auction schedule that maintains a high price until a sharp final drop is most effective in aggregating demand.

8. Abhay Sobhanan, Changhyun Kwon, Hadi Charkhgard. Arc Routing Problems with Multiple Trucks and Drones: A Hybrid Genetic Algorithm arXiv abstract

   Arc-routing problems underpin numerous critical field operations, including power-line inspection, urban police patrolling, and traffic monitoring. In this domain, the Rural Postman Problem (RPP) is a fundamental variant in which a prescribed subset of edges or arcs in a network must be traversed. This paper investigates a generalized form of the RPP, called RPP-mTD, which involves a fleet of multiple trucks, each carrying multiple drones. The trucks act as mobile depots traversing a road network, from which drones are launched to execute simultaneous service, with the objective of minimizing the overall makespan. Given the combinatorial complexity of RPP-mTD, we propose a Hybrid Genetic Algorithm (HGA) that combines population-based exploration with targeted neighborhood searches. Solutions are encoded using a two-layer chromosome that represents: (i) an ordered, directed sequence of required edges, and (ii) their assignment to vehicles. A tailored segment-preserving crossover operator is introduced, along with multiple local search techniques to intensify the optimization. We benchmark the proposed HGA against established single truck-and-drone instances, demonstrating competitive performance. Additionally, we conduct extensive evaluations on new, larger-scale instances to demonstrate scalability. Our findings highlight the operational benefits of closely integrated truck-drone fleets, affirming the HGA’s practical effectiveness as a decision-support tool in advanced mixed-fleet logistics.

9. Jae Hyeok Lee, Minwoo Kim, Minjun Kim, Jinkyoo Park, Changhyun Kwon. The Iterative Chainlet Partitioning Algorithm for the Traveling Salesman Problem with Drone and Neural Acceleration arXiv github abstract

   This study introduces the Iterative Chainlet Partitioning (ICP) algorithm and its neural acceleration for solving the Traveling Salesman Problem with Drone (TSP-D). The proposed ICP algorithm decomposes a TSP-D solution into smaller segments called chainlets, each optimized individually by a dynamic programming subroutine. The chainlet with the highest improvement is updated, and the procedure is repeated until no further improvement is possible. We show that the subroutine runs in quadratic time and the number of subroutine calls is bounded linearly in problem size for the first iteration and remains constant in subsequent iterations, ensuring algorithmic scalability. Empirical results show that ICP outperforms existing algorithms in both solution quality and computational time. Tested over 1,249 benchmark instances, ICP yields an average improvement of 2.6% in solution quality over the previous state-of-the-art algorithm while reducing computational time by 91.3%. The procedure is deterministic, ensuring reliability without requiring multiple runs. The subroutine is the computational bottleneck in the already efficient ICP algorithm. To reduce the necessity of subroutine calls, we integrate a graph neural network (GNN) to predict incremental improvements. We demonstrate that the resulting Neuro ICP (NICP) achieves substantial acceleration while maintaining solution quality. Compared to ICP, NICP reduces the total computational time by 28.6%, while the objective function value increase is limited to 0.14%. A transfer learning framework enables efficient extension to various operational constraints, making this a valuable foundation for developing efficient algorithms for truck-drone synchronized routing problems.

10. Sasan Mahmoudinazlou, Changhyun Kwon, Hadi Charkhgard. Solving Various Classes of Arc Routing Problems with a Memetic Algorithm-based Framework url abstract

    Arc routing problems are combinatorial optimization problems that have many real-world applications, such as mail delivery, snow plowing, and waste collection. Various variants of this problem are available, as well as algorithms intended to solve them heuristically or exactly. Presented here is a generic algorithmic framework that can be applied to a variety of arc routing problems where a fleet of vehicles is used to visit a predefined set of edges. The main characteristic of the problem that qualifies it for the proposed framework is that each edge should be visited no more than once. This proposed framework uses genetic algorithms, dynamic programming, and local searches in a systematic manner and provides guidelines for applying them to the arc routing problem of one’s choice. We select two problems to test the effectiveness of our proposed framework: the min-max windy K-vehicle rural postman problem and the undirected capacitated arc routing problem. We implement our proposed framework and compare it with existing algorithms using an established benchmark set for each problem. We demonstrate that our generic proposed framework can outperform existing custom-built algorithms.

Refereed Publications in Journals

1. Yoonju Sim, Federico Berto, Chuanbo Hua, Jinkyoo Park, Changhyun Kwon. LLM-Driven Algorithm Design for Quantum Circuit Synthesis based on Binary Decision Diagrams Transactions on Machine Learning Research, accepted. arXiv abstract

   Quantum circuits are central to implementing quantum algorithms on quantum devices, where quantum gates must be reversible. Many quantum algorithms rely on Boolean functions, which must therefore be implemented reversibly within quantum circuits. Reversible circuit synthesis provides a way to translate such Boolean functions into reversible circuits. Binary decision diagrams (BDDs) offer a scalable approach to this task, but the resulting BDDs and circuits depend heavily on variable ordering. Existing ordering heuristics commonly minimize BDD size because it is closely tied to the circuit size. However, BDD size is an imperfect proxy for the quantum cost of the synthesized circuit (QCC). We propose QuantumEvo, an evolutionary framework that uses an LLM as a heuristic generator for QCC-aware BDD variable ordering. Instead of predicting orderings directly, QuantumEvo searches over ordering heuristics initialized from multiple heuristic families. Candidate heuristics directly manipulate variable orderings using standard BDD operations and are selected by downstream QCC. The discovered heuristic, HGA-QE, modifies the sifting step inside a genetic algorithm so that the procedure is better aligned with QCC. Across the benchmark set, HGA-QE achieves a 70.9% tie-or-win rate against the per-function best baseline and is strictly best on 13.5% of the functions. The results demonstrate broadly competitive QCC performance, with HGA-QE showing a clearer relative advantage in strict wins on the two benchmark suites drawn from sources different from the data used for heuristic discovery

2. Youngjun Park, Jihye Na, Keonhee Jang, Dongman Lee, Changhyun Kwon, Yoonjin Yoon. Food Deserts in Context: How Site-Specific Factors Reshape Food Deserts Discussion. npj Urban Sustainability, accepted. pdf ssrn doi abstract

   Research on food deserts has evolved from a narrow focus on geographical distance to a comprehensive framework considering socioeconomic, infrastructural, and cultural factors. This review synthesizes global literature to examine how local contexts reshape food desert definitions, measurement, and interventions. We find a clear shift from distance-based metrics to multidimensional analyses incorporating affordability, mobility, and local food environments. Correspondingly, interventions are diverse, spanning from supermarket investments to digital platforms, with effectiveness highly dependent on site-specific conditions like urban density or social capital. We conclude that addressing food deserts demands integrated, context-sensitive policies, as a one-size-fits-all approach is insufficient. Future research should leverage mobility data, online food access trends, and localized models to build more effective food security strategies.

3. Xufei Liu, Sung Hoon Chung, Changhyun Kwon (2025) An Adaptive Large Neighborhood Search Method for the Drone-Truck Arc Routing Problem. Computers & Operations Research, 176, 106959. pdf doi abstract

   For applications such as traffic monitoring, infrastructure inspection, and security, ground vehicles (trucks) and unmanned aerial vehicles (drones) may collaborate to finish the task more efficiently. This paper considers an Arc Routing Problem (ARP) with a mixed fleet of a single truck and multiple homogeneous drones, called a Drone-Truck Arc Routing Problem (DT-ARP). While the truck must follow a road network, the drone can fly off of it. With a limited battery capacity, however, the drone has a length constraint, i.e., the maximum flight range. A truck driver can replace a battery for the drone after each flight trip. We first transform the DT-ARP into a node routing problem, for which we present a MIP formulation for the case with a truck and a drone. To solve large-size instances with multiple drones, a heuristic method based on Adaptive Large Neighborhood Search (ALNS) is proposed. The performance of ALNS is evaluated on small-size randomly generated instances and large-size undirected rural postman problem benchmark instances. In addition, an analysis is provided on the relationship between truck/drone speeds and the drone’s flight range, which affects the difficulty level to solve. The robustness of ALNS is shown via numerical experiments.

4. Abhay Sobhanan, Junyoung Park, Jinkyoo Park, Changhyun Kwon (2025) Genetic Algorithms with Neural Cost Predictor for Solving Hierarchical Vehicle Routing Problems. Transportation Science, 59(2), 207–450. arXiv doi github abstract

   When vehicle routing decisions are intertwined with higher-level decisions, the resulting optimization problems pose significant challenges for computation. Examples are the multi-depot vehicle routing problem (MDVRP), where customers are assigned to depots before delivery, and the capacitated location routing problem (CLRP), where the locations of depots should be determined first. A simple and straightforward approach for such hierarchical problems would be to separate the higher-level decisions from the complicated vehicle routing decisions. For each higher-level decision candidate, we may evaluate the underlying vehicle routing problems to assess the candidate. As this approach requires solving vehicle routing problems multiple times, it has been regarded as impractical in most cases. We propose a novel deep-learning-based approach called Genetic Algorithm with Neural Cost Predictor (GANCP) to tackle the challenge and simplify algorithm developments. For each higher-level decision candidate, we predict the objective function values of the underlying vehicle routing problems using a pre-trained graph neural network without actually solving the routing problems. In particular, our proposed neural network learns the objective values of the HGS-CVRP open-source package that solves capacitated vehicle routing problems. Our numerical experiments show that this simplified approach is effective and efficient in generating high-quality solutions for both MDVRP and CLRP and has the potential to expedite algorithm developments for complicated hierarchical problems. We provide computational results evaluated in the standard benchmark instances used in the literature.

5. Sasan Mahmoudinazlou, Changhyun Kwon (2024) A Hybrid Genetic Algorithm with Type-Aware Chromosomes for Traveling Salesman Problems with Drone. European Journal of Operational Research, 318(3), pp. 719–739. arXiv doi github abstract

   There are emerging transportation problems known as the Traveling Salesman Problem with Drone (TSPD) and the Flying Sidekick Traveling Salesman Problem (FSTSP) that involve using a drone in conjunction with a truck for package delivery. This study presents a hybrid genetic algorithm for solving TSPD and FSTSP by incorporating local search and dynamic programming. Similar algorithms exist in the literature. Our algorithm, however, considers more sophisticated chromosomes and less computationally complex dynamic programming to enable broader exploration by the genetic algorithm and efficient exploitation through dynamic programming and local search. The key contribution of this paper is the discovery of how decision-making processes for solving TSPD and FSTSP should be divided among the layers of genetic algorithm, dynamic programming, and local search. In particular, our genetic algorithm generates the truck and the drone sequences separately and encodes them in a type-aware chromosome, wherein each customer is assigned to either the truck or the drone. We apply local search to each chromosome, which is decoded by dynamic programming for fitness evaluation. Our new algorithm is shown to outperform existing algorithms on most benchmark instances in both quality and time. Our algorithms found the new best solutions for 538 TSPD instances out of 920 and 74 FSTSP instances out of 132.

6. Hyeonah Kim, Jinkyoo Park, Changhyun Kwon (2024) A Neural Separation Algorithm for the Rounded Capacity Inequalities. INFORMS Journal on Computing, 36(4), pp. 987–1005. (a featured article in the July/August 2024 issue of IJOC) arXiv doi github abstract

   The cutting plane method is a key technique for successful branch-and-cut and branch-price-and-cut algorithms that find the exact optimal solutions for various vehicle routing problems (VRPs). Among various cuts, the rounded capacity inequalities (RCIs) are the most fundamental. To generate RCIs, we need to solve the separation problem, whose exact solution takes a long time to obtain; therefore, heuristic methods are widely used. We design a learning-based separation heuristic algorithm with graph coarsening that learns the solutions of the exact separation problem with a graph neural network (GNN), which is trained with small instances of 50 to 100 customers. We embed our separation algorithm within the cutting plane method to find a lower bound for the capacitated VRP (CVRP) with up to 1,000 customers. We compare the performance of our approach with CVRPSEP, a popular separation software package for various cuts used in solving VRPs. Our computational results show that our approach finds better lower bounds than CVRPSEP for large-scale problems with more than 300 customers, while CVRPSEP shows strong competency for problems with less than 300 customers.

7. Sasan Mahmoudinazlou, Changhyun Kwon (2024) A Hybrid Genetic Algorithm for the min-max Multiple Traveling Salesman Problem. Computers & Operations Research, 162, 106455. arXiv doi github abstract

   This paper proposes a hybrid genetic algorithm for solving the Multiple Traveling Salesman Problem (mTSP) to minimize the length of the longest tour. The genetic algorithm utilizes a TSP sequence as the representation of each individual, and a dynamic programming algorithm is employed to evaluate the individual and find the optimal mTSP solution for the given sequence of cities. A novel crossover operator is designed to combine similar tours from two parents and offers great diversity for the population. For some of the generated offspring, we detect and remove intersections between tours to obtain a solution with no intersections. This is particularly useful for the min-max mTSP. The generated offspring are also improved by a self-adaptive random local search and a thorough neighborhood search. Our algorithm outperforms all existing algorithms on average, with similar cutoff time thresholds, when tested against multiple benchmark sets found in the literature. Additionally, we improve the best-known solutions for 21 out of 89 instances on four benchmark sets.

8. Xufei Liu, Sang Won Kim, Changhyun Kwon (2023) An Adaptive Large Neighborhood Search Method for Rebalancing Free-Floating Electric Vehicle Sharing Systems. Computers & Operations Research, 155, 106220. pdf doi abstract

   Free-floating electric vehicle sharing systems allow users to pick up an available electric vehicle (EV) and return it to any permissible parking location within a service area. Such service flexibility can drive a severe spatial imbalance between vehicle availability and trip demands. Hence, it is an important part of their operations to relocate the EV fleet to meet the next day’s demand with sufficient battery levels. This relocation operation involves a complicated routing problem for a fleet of shuttles to transport the staff drivers who recharge, if necessary, and relocate the EVs to proper demand locations. Characterized by unique hierarchical and interdependent decisions, the EV relocation and shuttle routing problem poses significant computational challenges for large-scale problems. We devise an efficient algorithm that adapts the Adaptive Large Neighborhood Search (ALNS) metaheuristic framework to overcome such unique challenges. The algorithm is tested on two sets of data: randomly generated data and real-world EV-sharing usage data in Amsterdam. The results validate the efficiency and effectiveness of our ALNS algorithm. In addition, the analysis of total operational cost and waiting time percentage provides practical recommendations to decision-makers on choosing the mode of staff transportation (e.g., shuttles vs. personal mobility such as scooters). Lastly, our numerical results also highlight the usefulness of our ALNS method, which is quite flexible to be applied to a dynamic environment where some EV demands are removed or added in the course of EV relocation operations.

9. Aigerim Bogyrbayeva, Taehyun Yoon, Hanbum Ko, Sungbin Lim, Hyokun Yun, Changhyun Kwon (2023) A Deep Reinforcement Learning Approach for Solving the Traveling Salesman Problem with Drone. Transportation Research Part C: Emerging Technologies, 148, 103981 arXiv doi github abstract

   Reinforcement learning has recently shown promise in learning quality solutions in many combinatorial optimization problems. In particular, the attention-based encoder-decoder models show high effectiveness on various routing problems, including the Traveling Salesman Problem (TSP). Unfortunately, they perform poorly for the TSP with Drone (TSP-D), requiring routing a heterogeneous fleet of vehicles in coordination—a truck and a drone. In TSP-D, the two vehicles are moving in tandem and may need to wait at a node for the other vehicle to join. State-less attention-based decoder fails to make such coordination between vehicles. We propose a hybrid model that uses an attention encoder and a Long Short-Term Memory (LSTM) network decoder, in which the decoder’s hidden state can represent the sequence of actions made. We empirically demonstrate that such a hybrid model improves upon a purely attention-based model for both solution quality and computational efficiency. Our experiments on the min-max Capacitated Vehicle Routing Problem (mmCVRP) also confirm that the hybrid model is more suitable for the coordinated routing of multiple vehicles than the attention-based model. The proposed model demonstrates comparable results as the operations research baseline methods.

10. Zulqarnain Haider, Yujie Hu, Hadi Charkhgard, David Himmelgreen, Changhyun Kwon (2022) Creating Grocery Delivery Hubs for Food Deserts at Local Convenience Stores via Spatial and Temporal Consolidation. Socio-Economic Planning Sciences, 82(B), 101301. doi pdf abstract

    For many socioeconomically disadvantaged customers living in food deserts, the high costs and minimum order size requirements make attended grocery deliveries financially non-viable, although it has potential to provide healthy foods to the food insecure population. This paper proposes consolidating customer orders and delivering to a neighborhood convenience store instead of home delivery. We employ an optimization framework involving the minimum cost set covering and the capacitated vehicle routing problems. Our experimental studies in three counties in the U.S. suggest that by spatial and temporal consolidation of orders, the deliverer can remove minimum order-size requirements and reduce the delivery costs, depending on various factors, compared to attended home-delivery. We find the number and size of time windows for home delivery to be the most important factor in achieving temporal consolidation benefits. Other significant factors in achieving spatial consolidation include the capacity of delivery vehicles, the number of depots, and the number of customer orders. We also find that the number of partner convenience stores and the walkable distance parameter of the model, have a significant impact on the number of accepted orders, i.e., the service level provided by the deliverer. The findings of this study imply consolidated grocery delivery as a viable solution to improve fresh food access in food deserts. In light of the recent global pandemic, and its exacerbating effects on food insecurity, the innovative solution proposed in this paper is even more relevant and timely.

11. Aigerim Bogyrbayeva, Sungwook Jang, Ankit Shah, Young Jae Jang, Changhyun Kwon (2020) A Reinforcement Learning Approach for Rebalancing Electric Vehicle Sharing Systems. IEEE Transactions on Intelligent Transportation Systems, 23(7), 8704–8714. doi arXiv pdf github abstract

    This paper proposes a reinforcement learning approach for nightly offline rebalancing operations in free-floating electric vehicle sharing systems (FFEVSS). Due to sparse demand in a network, FFEVSS require relocation of electrical vehicles (EVs) to charging stations and demander nodes, which is typically done by a group of drivers. A shuttle is used to pick up and drop off drivers throughout the network. The objective of this study is to solve the shuttle routing problem to finish the rebalancing work in the minimal time. We consider a reinforcement learning framework for the problem, in which a central controller determines the routing policies of a fleet of multiple shuttles. We deploy a policy gradient method for training recurrent neural networks and compare the obtained policy results with heuristic solutions. Our numerical studies show that unlike the existing solutions in the literature, the proposed methods allow to solve the general version of the problem with no restrictions on the urban EV network structure and charging requirements of EVs. Moreover, the learned policies offer a wide range of flexibility resulting in a significant reduction in the time needed to rebalance the network.

12. Aigerim Bogyrbayeva, Changhyun Kwon (2021) Pessimistic Evasive Flow Capturing Problems. European Journal of Operational Research, 293(1), 133-148. doi pdf abstract

    The evasive flow capturing problem (EFCP) is to locate a set of law enforcement facilities to intercept unlawful flows. One application of the EFCP is the location problem of weight-in-motion systems deployed by authorities to detect overloaded vehicles characterized by evasive behavior. Although optimistic bilevel formulations of the EFCP are studied in the literature, we provide two pessimistic formulations by relaxing the assumption on the optimizing behavior of unlawful travelers and capturing the ambiguity of travelers’ route choices. While the first pessimistic formulation yields the most robust network design, the second pessimistic formulation can flexibly control the level of pessimism on the behavior of followers between optimistic and overly pessimistic. The resulting formulations yield a robust network design and represent realistic behavior of drivers. The pessimistic formulations introduce another level in the optimization problem, for which we propose a cutting plane algorithm. The proposed solution methods demonstrate their effectiveness on real and randomly generated networks. We also compute the value of pessimistic solutions.

13. Aigerim Bogyrbayeva, Mahdi Takalloo, Hadi Charkhgard, Changhyun Kwon (2021) An Iterative Combinatorial Auction Design for Fractional Ownership of Autonomous Vehicles. International Transactions in Operational Research, 28(4), 1681–1705. doi pdf abstract

    This study designs a new market for fractional ownership of autonomous vehicles (AVs), in which an AV is co-leased by a group of individuals. We present a practical iterative auction based on the combinatorial clock auction to match the interested customers together and determine their payments. In designing such an auction, we consider continuous-time items (time slots) which are defined by bidders and naturally exploit driverless mobility of AVs to form co-leasing groups. To relieve the computational burdens of both bidders and the auctioneer, we devise user agents who generate packages and bid on behalf of bidders. Through numerical experiments using the California 2010–2012 travel survey, we test the performance of the auction design. We also compare various bidding strategies and study the effect of activity rules on the bidders’ payoffs. We find that the designed activity rules successfully remove the strategic behavior of bidders. We also find that core-selecting payment rule brings the largest revenue to the auctioneer in most cases.

14. Mahdi Takalloo, Aigerim Bogyrbayeva, Hadi Charkhgard, Changhyun Kwon (2021) Solving the Winner Determination Problem in Combinatorial Auctions for Fractional Ownership of Autonomous Vehicles. International Transactions in Operational Research, 28(4), 1658–1680. doi pdf Fractional Ownership; Combinatorial Auction; Bidder-Defined Items; Clique abstract

    The introduction of autonomous vehicles to consumer markets will expedite the trend of car-sharing and enable co-owning or co-leasing a car. In this paper, we consider a combinatorial auction market for fractional ownership of autonomous vehicles, which is unique in two aspects. First, items are neither pre-defined nor discrete; rather, items are continuous time slots defined by bidders. Second, the spatial information of bidders should be incorporated within the winner determination problem so that sharing a vehicle is indeed a viable plan. The consideration of spatial information increases the computational complexity significantly. We formulate the winner determination problem, which plays a critical role in various auction designs and pricing schemes, for both discrete- and continuous-time settings. In terms of social welfare maximization, we show that the continuous-time model is superior to the discrete-time model. We provide a conflict-based reformulation of the continuous-time model, for which we develop an effective solution approach based on a heuristic and maximal-clique based reformulations. Using samples of the 2010–2012 California Household Travel Survey, we verify that the proposed solution methods provide effective computational tools for the combinatorial auction with bidder-defined items.

15. Mahdi Takalloo, Changhyun Kwon (2020) On the Price of Satisficing in Network User Equilibria. Transportation Science, 54(6), 1555–1570. doi arXiv pdf bounded rationality; satisficing; user equilibrium; sensitivity analysis abstract

    When drivers are satisficing decision-makers, the resulting traffic pattern attains a satisficing user equilibrium, which may deviate from the (perfectly rational) user equilibrium. In a satisficing user equilibrium traffic pattern, the total system travel time can be worse than in the case of the PRUE. We show how bad the worst-case satisficing user equilibrium traffic pattern can be, compared to the perfectly rational user equilibrium. We call the ratio between the total system travel times of the two traffic patterns the price of satisficing, for which we provide an analytical bound. Using the sensitivity analysis for variational inequalities, we propose a numerical method to quantify the price of satisficing for any given network instance.

16. Kevin A. Melendez, Tapas K. Das, Changhyun Kwon (2020) A Nash-bargaining Model for Trading of Electricity Between Aggregations of Peers. International Journal of Electrical Power and Energy Systems, 123, 106185. doi pdf Aggregation of peers; peer-to-peer energy trading; option contract; Nash bargaining solution abstract

    In the last several years, the growth in household solar generation and the lack of success of the feed-in-tariff programs have led to the rise of peer-to-peer (P2P) energy trading schemes among prosumers. However, a change that has started more recently is the growth of smart homes and businesses, of which loads are IoT controlled and are supported by advanced metering infrastructure (AMI). This has created a new opportunity for smart homes and businesses to form aggregations (coalitions) and participate in cooperative load management and energy trading. Unlike energy trading among individual prosumers in most P2P networks, a new trading opportunity that is emerging is between aggregations of peers of smart homes and businesses and electric vehicles (EVs). In this paper, we consider one such trading scenario between two aggregations, of which one has smart homes and businesses with load consuming entities (not prosumers), and the other has EVs only. The aggregation with smart homes and businesses derive cost reduction through optimal load scheduling based on load preferences, market-based pricing of electricity, and opportunity to trade (buy) energy from the aggregation with EVs. Whereas the aggregation of EVs optimally schedules charging to meet EV needs and uses stored energy to trade (sell). A generalized Nash bargaining model is developed for obtaining optimal trading strategies in the form of plain or swing option contracts. A sample numerical problem scenario is used to show that suitable contracts can be derived that allow aggregations of peers to mutually benefit from energy trading. Interactions among contract parameters (such as strike price, option value, and option quantity) and the relative market power of the aggregations are also examined.

17. Kevin A. Melendez, Tapas K. Das, Changhyun Kwon (2020) Optimal operation of a system of charging hubs and a fleet of shared autonomous electric vehicles. Applied Energy, 279, 115861. doi pdf Shared autonomous electric vehicles, cyber-physical system, ACOPF, energy arbitrage, robust optimization, stochastic optimization abstract

    Shared autonomous electric vehicles (SAEVs) are expected to serve a significant fraction of the passenger transportation needs in cities and surrounding urban areas. In this paper, we consider optimal operation of a cyber-physical system (CPS) comprising a large fleet of SAEVs and a set of charging hubs located across the transportation network and supported by the power grid. The hubs are considered to have a number of charging stations, a stand-alone battery bank for energy storage, and limited rooftop photo-voltaic (PV) generation capacity. We developed a robust mixed integer linear programming model. It considers a number of practical features of both the power and transportation systems, including day-ahead load commitment for electricity via an alternative current power flow model, real time price spikes of electricity, energy arbitrage, uncertainty in passenger demand, and balking of passengers while waiting for a ride. We demonstrated our methodology by implementing it on a sample CPS with 500 SAEVs and five hubs with fifty charging stations in each. Our methodology yields operational decisions for day ahead commitment of power and real time control of the SAEVs and the hubs. The sample CPS is used to examine impact of hub capacity and fleet size on various system performance measures. We discuss the computational challenges of our methodology and propose a simplified myopic approach that is capable of dealing with much larger fleet sizes and a variety of hub capacities. Reduction in computation time and the optimality gap for the myopic approach are examined.

18. Liu Su, Changhyun Kwon (2020) Risk-Averse Network Design with Behavioral Conditional Value-at-Risk for Hazardous Materials Transportation. Transportation Science, 54(1), 184–203. doi pdf transportation; hazardous materials; network design; conditional value-at-risk; Benders decomposition abstract

    We consider a road-ban problem in hazardous materials (hazmat) transportation. We formulate the problem as a network design problem to select a set of closed road segments for hazmat traffic and obtain a bi-level optimization problem. While modeling probabilistic route-choices of hazmat carriers by the random utility model (RUM) in the lower level, we consider a risk-averse measure called conditional value-at-risk (CVaR) in the upper level, instead of the widely used expected risk measure. Using RUM and CVaR, we quantify the risk of having hazmat accidents and large consequences, and design the network policy for road-bans accordingly. While CVaR has been used in hazmat routing problems, this paper is the first attempt to apply CVaR in risk averse hazmat network design problems considering stochastic route-choices of hazmat carriers. The resulting problem is a mixed integer nonlinear programming problem, for which we devise a line search approach combined with Benders decomposition. We demonstrate the efficiency of the proposed computational method with case studies. The average computation time for a network with 105 nodes and 268 arcs is 3 hours. Commercial solvers are inadequate to solve this problem, because the optimality gap is 99.9% after 24 hours just for a linear subproblem. By applying CVaR to the route-choice behavior of hazmat carriers, we protect the road network from undesirable route-choices that may lead to severe consequences. We define the Value of RUM-CVaR Solutions (VRCS) over the deterministic model based on shortest-path problems and the expected risk measure. Our case study shows that VRCS can range from 4.9% to 64.1% depending on the probability threshold used in the CVaR measure.

19. Anpeng Zhang, Jee Eun Kang, Changhyun Kwon (2020) Generalized Stable User Matching for Autonomous Vehicle Co-ownership Programs. Service Science, 12(2–3), 61–79. doi pdf autonomous vehicles; stable matching; roommate matching; car sharing system; vehicle ownership; fractional ownership abstract

    We investigate a new form of car sharing system that can be introduced in the market for autonomous vehicles, called fractional ownership or co-ownership. While dynamic ride sharing provides ad-hoc shared mobility services without any long-term commitment, we consider co-ownership programs with which users can still “own” a car with committed usages and ownership. We assume that an autonomous vehicle is shared by a group of users, which is only accessible by the group. We use stable matching to help users find an appropriate group to share an autonomous vehicle and present a generalized stable matching model that allows flexible sizes of groups as well as various alternative objectives. We also present a heuristic algorithm to improve computational time due to the combinatorial property of the problem.

20. Anpeng Zhang, Jee Eun Kang, Changhyun Kwon (2020) Multi-day Scenario Analysis for Battery Electric Vehicle Feasibility Assessment and Charging Infrastructure Planning. Transportation Research Part C: Emerging Technologies, 111, 439–457. doi pdf activity-travel patterns; battery electric vehicle; charging infrastructure planning; level 3 charging abstract

    Multi-day activity-travel patterns help create potential vehicle usage profiles that contain vehicle operations and battery status under different scenarios with varying location-based charging opportunities, based on travel needs and charging availability/behaviors. Utilizing a multi-day data sampling method, analyses of scenarios are designed to provide insights on bounds of potential BEV market under different charging opportunities, including level 2 activity charging and level 3 trip charging. Single-day data results tend to overestimate travelers’ BEV feasibility assuming that multi-day sample data provides accurate estimations. Facility utilization can be improved without affecting travelers’ charging demand under correct pricing scheme for most cost-sensitive users. Smart grid charging strategy can greatly reduce the total number of operating chargers during the same time in a day, and BEV users’ charging behaviors have minor impact on this improvement. Our numerical results indicate that an appropriate number of chargers installed in shopping and leisure locations should be more profitable and have higher charger utilization rate since those chargers help cover BEV users’ trips.

21. Masoumeh Taslimi, Rajan Batta, Changhyun Kwon (2020) Medical Waste Collection Considering Transportation and Storage Risk. Computers & Operations Research, 120, 104966. doi pdf medical waste collection; hazardous materials transportation; vehicle routing; decomposition-based heuristic abstract

    We consider a Periodic Load-dependent Capacitated Vehicle Routing Problem (PLCVRP) encountered by healthcare centers and medical waste collection companies for the design of a weekly inventory routing schedule to transport medical wastes to treatment sites. In addition to minimization of transportation risk, occupational risk related to temporary storage of hazardous wastes at the healthcare centers is considered. The transport risk on each arc is dependent on the weight of hazardous medical waste on the vehicle when it traverses that arc. We devise a decomposition based heuristic algorithm to solve this problem. We analyze the characteristics of the PLCVRP’s solutions with respect to four different criteria: (i) transport and occupational risk, (ii) transport risk, (iii) occupational risk, and (iv) transportation cost. Solving different versions of PLCVRP reveals that minimizing both transport and occupational risk on the network can aid decision makers to develop a better routing schedule in terms of the imposed risk of hazardous medical waste. Experimental results confirm the efficiency of our heuristic. We present a case study to illustrate solution attributes obtained by our solution methodology. The case study is based on medical waste management in Dolj, Romania.

22. Mahdi Takalloo, Changhyun Kwon (2020) Sensitivity of Wardrop Equilibria: Revisited. Optimization Letters, 14, 781–796. doi arXiv pdf Wardrop equilibria; Selfish routing; Sensitivity analysis abstract

    For single-commodity networks, the increase of the price of anarchy is bounded by a factor of (1 + ε)^p from above, when the travel demand is increased by a factor of 1 + ε and the latency functions are polynomials of degree at most p. We show that the same upper bound holds for multi-commodity networks and provide a lower bound as well.

23. Xufei Liu, Changhyun Kwon (2020) Exact Robust Solutions for the Combined Facility Location and Network Design Problem in Hazardous Materials Transportation. IISE Transactions, 52(10), 1156–1172. (Runner-up, Student Paper Competition of the INFORMS Section on Location Analysis (SOLA), 2019) doi pdf bi-level optimization; facility location; network design; cutting plane; Benders decomposition; hazardous materials abstract

    We consider a leader-follower game in the form of a bi-level optimization problem that simultaneously optimizes facility locations and network design in hazardous materials transportation. In the upper level, the leader intends to reduce the facility setup cost and the hazmat exposure risk, by choosing facility locations and road segments to close for hazmat transportation. When making such decisions, the leader anticipates the response of the followers who want to minimize the transportation costs. Considering uncertainty in the hazmat exposure and the hazmat transport demand, we consider a robust optimization approach with multiplicative uncertain parameters and polyhedral uncertainty sets. The resulting problem has a min-max problem in the upper level and a shortest-path problem in the lower level. We devise an exact algorithm that combines a cutting plane algorithm with Benders decomposition.

24. Mitchell J. Eaton, Simeon Yurek, Zulqarnain Haider, Julien Martin, Fred A. Johnson, Bradley J. Udell, Hadi Charkhgard, Changhyun Kwon (2019) Spatial Conservation Planning under Uncertainty: Adapting to Climate Change Risks using Modern Portfolio Theory. Ecological Applications, 29(7):e01962. doi Reserve design, spatial conservation planning, modern portfolio theory, multi-criteria decision analysis, risk management, sea-level rise, urbanization, climate uncertainty abstract

    Climate change and urban growth impact habitats, species, and ecosystem services. To buffer against global change, an established adaptation strategy is designing protected areas to increase representation and complementarity of biodiversity features. Uncertainty regarding the scale and magnitude of landscape change complicates reserve planning and exposes decision makers to risk of failing to meet conservation goals. Conservation planning tends to treat risk as an absolute measure, ignoring the context of the management problem and risk preferences of stakeholders. Application to conservation of risk management theory emphasizes diversification of portfolio of assets, with the goal of reducing the impact of system volatility on investment return. We use principles of Modern Portfolio Theory (MPT), which quantifies risk as the variance and correlation among assets, to formalize diversification as an explicit strategy for managing risk in climate-driven reserve design. We extend MPT to specify a framework that evaluates multiple conservation objectives, allows decision makers to balance management benefits and risk when preferences are contested or unknown, and includes additional decision options such as parcel divestment when evaluating candidate reserve designs. We apply an efficient search algorithm that optimizes portfolio design for large conservation problems and a game theoretic approach to evaluate portfolio tradeoffs that satisfy decision makers with divergent benefit and risk tolerances, or when a single decision maker cannot resolve their own preferences. Evaluating several risk profiles for a case study in South Carolina, our results suggest that a reserve design may be somewhat robust to differences in risk attitude but that budgets will likely be important determinants of conservation planning strategies, particularly when divestment is considered a viable alternative. We identify a possible fiscal threshold where adequate resources allow protecting a sufficiently diverse portfolio of habitats such that the risk of failing to achieve conservation objectives is considerably lower. For a range of sea-level rise projections, conversion of habitat to open water (14--180%) and wetland loss (1–7%) are unable to be compensated under the current protected network. In contrast, optimal reserve design outcomes are predicted to ameliorate expected losses relative to current and future habitat protected under the existing conservation estate.

25. Vignesh Subramanian, Tapas K. Das, Changhyun Kwon, Abhijit Gosavi (2019) A Data-Driven Methodology for Dynamic Pricing and Demand Response in Electric Power Networks. Electric Power Systems Research, 174, 105869. doi pdf dynamic pricing; aggregated demand response; electric power network; Bayesian demand prediction abstract

    The practice of disclosing price of electricity before consumption (dynamic pricing) is essential to promote aggregator-based demand response in smart and connected communities. However, both practitioners and researchers have expressed fear that wild fluctuations in demand response resulting from dynamic pricing may adversely affect the stability of both the network and the market. This paper presents a comprehensive methodology guided by a data-driven learning model to develop stable and coordinated strategies for both dynamic pricing as well as demand response. The methodology is designed to learn offline without interfering with network operations. Application of the methodology is demonstrated using simulation results from a sample 5-bus PJM network. Results show that it is possible to arrive at stable dynamic pricing and demand response strategies that can reduce cost to the consumers as well as improve network load balance.

26. Kevin A. Melendez, Vignesh Subramanian, Tapas K. Das, Changhyun Kwon (2019) Empowering end-use consumers of electricity to aggregate for demand-side participation. Applied Energy, 248, 372–382. doi Demand-side participation; Aggregation of end-use consumers; Fairness; Peer-to-peer trading; Energy sharing using EVs; Optimization of power systems abstract

    End-use consumers (peers) are being empowered to aggregate for direct demand-side participation through load scheduling and energy sharing. This is the result of the growth of Internet of Things (IoT) enabled loads, availability of advanced metering infrastructure, and the move towards real-time (RT) pricing of electricity. Peer-to-peer (P2P) cooperation has received significant interest in recent years, though the focus of this growing body of research is on modeling prosumer behavior in microgrids. Hence, there is a need for new methodologies to examine empowerment of all end-use consumers (not limited to prosumers) to form aggregations and develop fair rules of cooperation to reduce cost. This paper offers an optimization based methodology to address the above need for power systems. It minimizes the total cost and considers fairness using a Nash bargaining approach. Since cost and fairness are often in conflict, trade-off strategies are also presented. The model to asses fairness is nonlinear. Hence, it is transformed into a second order cone program (SOCP) and solved using GUROBI software version 7.5.2. The methodology is implemented on a sample 5-bus network, built using price and demand data from one of the load zones of Pennsylvania, New Jersey, and Maryland (PJM) power network in the United States. It is shown that two aggregations of peers participating in the sample network can reduce their total cost by 14.17% and 22.7%, while maintaining fairness. Concluding remarks highlight some of the limitations of the methodology.

27. Longsheng Sun, Mark H. Karwan, Changhyun Kwon (2019) Path-Based Approaches to Robust Network Design Problems Considering Boundedly Rational Network Users. Transportation Research Record, 2673(3), 637–645. doi pdf networks and graphs; satisficing; network design; hazardous materials; bi-level optimization abstract

    Network users may choose non-shortest paths, when (1) they satisfice with sub-optimal routes, or (2) they have perception errors of the decision environment. The notion of generalized bounded rationality has been recently proposed to create a unified framework for these two sources of behavioral uncertainty in route choices. When the notion of generalized bounded rationality is used in robust network design problems, we obtain a bi-level optimization problem with the min-max objective function at the upper level, with three layers of optimization in total. In this paper, we derive equivalent single-level path-based formulations that are readily solvable by available optimization libraries. We show how to incorporate them into robust multi-commodity network design problems in hazardous materials transportation.

28. Liu Su, Longsheng Sun, Mark H. Karwan, Changhyun Kwon (2019) Spectral Risk Measure Minimization in Hazardous Materials Transportation. IISE Transactions, 59(6), 638–652. (Featured in IISE Magazine) doi pdf hazardous materials transportation; risk management; spectral risk; coherent risk measures abstract

    Due to catastrophic consequences of potential accidents in hazardous materials (hazmat) transportation, a risk-averse approach for routing is necessary. In this paper, we consider spectral risk measures, for risk-averse hazmat routing, which overcome challenges posed in the existing approaches such as conditional value-at-risk. In spectral risk measures, one can define the spectrum function precisely to reflect the decision maker’s risk preference. We show that spectral risk measures can provide a unified routing framework for popular existing hazmat routing methods based on expected risk, maximum risk, and conditional value-at-risk. We first consider a special class of spectral risk measures, for which the spectrum function is represented as a step function. We develop a mixed integer linear programming model in hazmat routing to minimize these special spectral risk measures and propose an efficient search algorithm to solve the problem. For general classes of spectral risk measures, we suggest approximation methods and path-based approaches. We propose an optimization procedure to approximate general spectrum functions using a step function. We illustrate the usage of spectral risk measures and the proposed computational approaches using data from real road networks.

29. Payman G. Saghand, Hadi Charkhgard, Changhyun Kwon (2019) A Branch-and-Bound Algorithm for a Class of Mixed Integer Linear Maximum Multiplicative Programs: A Multi-objective Optimization Approach. Computers & Operations Research, 101, 263–274. doi pdf Multiplicative programming; Multi-objective optimization; Optimization over the efficient set; Linear programming; Branch-and-bound algorithm abstract

    We present a linear programming based branch-and-bound algorithm for a class of mixed integer optimization problems with a bi-linear objective function and linear constraints. This class of optimization problems can be viewed as a special case of the problem of optimization over the set of efficient solutions in multi-objective optimization. It is known that when there exists no integer decision variable, such a problem can be solved in polynomial time. In fact, in such a case, the problem can be transformed into a Second-Order Cone Program (SOCP) and so it can be solved efficiently by a commercial solver such as CPLEX SOCP solver. However, in a recent study, it is shown that such a problem can be solved even faster in practice by using a bi-objective linear programming based algorithm. So, in this study, we embed that algorithm in an effective branch-and-bound framework to solve mixed integer instances. We also develop several enhancement techniques including preprocessing and cuts. An extensive computational study demonstrate that the proposed branch-and-bound algorithm outperforms a commercial mixed integer SOCP solver. Moreover, the effect of different branching and node selecting strategies is explored.

30. Zulqarnain Haider, Alexander Nikolaev, Jee Eun Kang, Changhyun Kwon (2018) Inventory Rebalancing through Pricing in Public Bike Sharing Systems. European Journal of Operational Research, 270(1), 103–117. doi pdf transportation; bike-sharing; shared-mobility; rebalancing; pricing; heuristics abstract

    This paper presents a new conceptual approach to improve the operational performance of public bike sharing systems using pricing schemes. Its methodological developments are accompanied by experimental analyses with bike demand data from Capital Bikeshare program of Washington, DC. An optimized price vector determines the incentive levels that can persuade system customers to take bicycles from, or park them at, neighboring stations so as to strategically minimize the number of imbalanced stations. This strategy intentionally makes some imbalanced stations even more imbalanced, creating hub stations. This reduces the need for trucks and dedicated staff to carry out inventory repositioning. For smaller networks, a bi-level optimization model is introduced to minimize the number of imbalanced stations optimally. The results are compared with a heuristic approach that adjusts route prices by segregating the stations into different categories based on their current inventory profile, projected future demand, and maximum and minimum inventory values calculated to fulfill certain desired service level requirements. We use a routing model for repositioning trucks to show that the proposed optimization model and the latter heuristic approach, called the iterative price adjustment scheme (IPAS), reduce the overall operating cost while partially or fully obviating the need for a manual repositioning operation.

31. Anpeng Zhang, Jee Eun Kang, Kay Axhausen, Changhyun Kwon (2018) Multi-day Activity-Travel Pattern Sampling Based on Single-Day Data. Transportation Research Part C: Emerging Technologies, 89, 96–112. doi pdf Activity-Travel Patterns; Day-to-day variability; Interpersonal variability; Sampling Multiday Activity-Travel Patterns abstract

    Although it is important to consider multi-day activities in transportation planning, multi-day activity-travel data are expensive to acquire and therefore rarely available. In this study, we propose to generate multi-day activity-travel data through sampling from readily available single-day household travel survey data. A key observation we make is that the distribution of interpersonal variability in single-day travel activity datasets is similar to the distribution of intrapersonal variability in multi-day. Thus, interpersonal variability observed in cross-sectional single-day data of a group of people can be used to generate the day-to-day intrapersonal variability. The proposed sampling method is based on activity-travel pattern type clustering, travel distance and variability distribution to extract such information from single-day data. Validation and stability tests of the proposed sampling methods are presented.

32. Zulqarnain Haider, Hadi Charkhgard, Changhyun Kwon (2018) A Robust Optimization Approach for Solving Problems in Conservation Planning. Ecological Modelling, 368, 288–297. doi pdf conservation planning; robust optimization; invasion control; reserve selection; bi-objective mixed integer linear programming abstract

    In conservation planning, the data related to size, growth and diffusion of populations is sparse, hard to collect and unreliable at best. If and when the data is readily available, it is not of sufficient quantity to construct a probability distribution. In such a scenario, applying deterministic or stochastic approaches to the problems in conservation planning either ignores the uncertainty completely or assumes a distribution that does not accurately describe the nature of uncertainty. To overcome these drawbacks, we propose a robust optimization approach to problems in conservation planning that considers the uncertainty in data without making any assumption about its probability distribution. We explore two of the basic formulations in conservation planning related to reserve selection and invasive species control to show the value of the proposed robust optimization. Several novel techniques are developed to compare the results produced by the proposed robust optimization approach and the existing deterministic approach. For the case when the robust optimization approach fails to find a feasible solution, a novel bi-objective optimization technique is developed to handle infeasibility by modifying the level of uncertainty. Some numerical experiments are conducted to demonstrate the efficacy of our proposed approach in finding more applicable conservation planning strategies.

33. Byung Do Chung, Sungjae Park, Changhyun Kwon (2018). Equitable Distribution of Recharging Stations for Electric Vehicles. Socio-Economic Planning Science, 63, 1–11. doi pdf electric vehicles; flow refueling location problem; demand equity; flow equity abstract

    Given the limited driving range of battery electric vehicles and lack of sufficient charging infrastructure, locating charging stations is an important decision problem to enable long-distance travels by battery electric vehicles. This paper considers an important political factor in such location problems: the equitable access to charging stations among geographical regions. We propose three types of equity constraints to the flow refueling location model: two constraints based on travel demand and the other based on flow. For solving the problem with flow equity constraints, we propose a multi-phase heuristic method. We test the proposed models and computational method in a real expressway network in Korea.

34. Longsheng Sun, Mark H. Karwan, Changhyun Kwon (2018). Generalized Bounded Rationality and Robust Multicommodity Network Design. Operations Research, 66(1), 42–57. (Outstanding Paper Award in Urban Transportation Planning and Modeling, the INFORMS Transportation Science & Logistics (TSL) Society, 2019) doi pdf bounded rationality; satisficing; perception; network design; robust optimization; inverse optimization abstract

    Often network users are not perfectly rational, especially when they are satisficing—rather than optimizing—decision makers and each individual’s perception of the decision environment reflects personal preferences or perception errors due to lack of information. While the assumption of satisficing drivers has been used in modeling route choice behavior, this research uses a link-based perception error model to describe driver’s uncertain behavior, without assuming stochasticity. In congestion-free networks, we show that the perception error model is more general than the existing bounded rationality models with satisficing drivers with special cases when the two approaches yield the same results; that is, satisficing under accurate perception is equivalent to optimizing under inaccurate perception. This motivates us to define generalized bounded rationality in route choice behavior modeling. The proposed modeling framework is general enough to capture link-specific cost-perception of drivers. We use a Monte Carlo method to estimate modeling parameter values to guarantee a certain coverage probability in comparison with the random utility model. We demonstrate how the notion of generalized bounded rationality can be used in robust multicommodity network design problems and devise a cutting plane algorithm. We illustrate our approaches in the context of hazardous materials transportation.

35. Tolou Esfandeh, Rajan Batta, Changhyun Kwon (2018). Time-Dependent Hazardous-materials Network Design Problem. Transportation Science, 52(2), 454–473. doi pdf github hazardous materials transportation; network design; column generation; label setting abstract

    We extend the hazardous-materials (hazmat) network design problem to account for the time-dependent road closure as a policy tool in order to reduce hazmat transport risk by altering carriers’ departure times and route choices. We formulate the time-dependent network design problem using an alternative-based model with each alternative representing a combined path and departure-time choice. We also present an extended model that can not only account for consecutive time-based road closure policies, but also allow stopping at the intermediate nodes of the network in the routing/scheduling decisions of the carriers. Heuristic algorithms based on column-generation and label-setting are presented. To illustrate the advantages that can be gained through the use of our methodology, we present results from numerical experiments based on a transportation network from Buffalo, NY. To investigate the impact of the extensions, we consider three versions of the problem by gradually refining the model. We show that under consideration of extensions, the design policies are more applicable and effective.

36. Iakovos Toumazis, Murat Kurt, Artemis Toumazi, Loukia G Karakosta, Changhyun Kwon (2017). Comparative Effectiveness of Up-to-Three Lines of Chemotherapy Treatment Plans for Metastatic Colorectal Cancer. Medical Decision Making doi pdf abstract

    Modern chemotherapy agents transformed standard care for metastatic colorectal cancer (mCRC) but raised concerns about the financial burden of the disease. We studied comparative effectiveness of treatment plans that involve up to three lines of therapies and impact of treatment sequencing on health and cost outcomes. We employed a Markov model to represent the dynamically changing health status of mCRC patients and used Monte-Carlo simulation to evaluate various treatment plans consistent with existing guidelines. We calibrated our model by a meta-analysis of published data from an extensive list of clinical trials and measured the effectiveness of each plan in terms of cost per quality-adjusted life-year (QALY). We examined the sensitivity of our model and results with respect to key parameters in two scenarios serving as base- and worst-cases for patients’ overall and progression free survivals. The derived efficient frontiers included 7 and 5 treatment plans in base- and worst-cases, respectively. The incremental cost-effectiveness ratio (ICER) ranged between $26,260 and $152,530 when the treatment plans on the efficient frontiers were compared against the least costly efficient plan in base-case, and between $21,256 to $60,040 in worst-case. All efficient plans were expected to lead to fewer than 2.5 AEs and on average successive AEs were spaced more than 9 weeks apart from each other in base-case. Based on ICER, all efficient treatment plans exhibit at least 87% chance of being efficient. Sensitivity analyses show that the ICERs were most dependent on drug acquisition cost, distributions of progression free and overall survivals, and health utilities. We conclude that improvements in health outcomes may come at high incremental costs and are highly dependent in the order treatments are administered.

37. Anpeng Zhang, Jee Eun Kang, Changhyun Kwon (2017). Incorporating Demand Dynamics in Multi-Period Capacitated Fast-charging Location Planning for Electric Vehicles. Transportation Research Part B: Methodological, 103, 5–29. doi pdf github electric vehicles; facility location; flow refueling location problem; multi-period planning; vehicle market demand dynamics; alternative fuel vehicles abstract

    We develop a multi-period capacitated flow refueling location problem for electric vehicles (EVs) as EV market responds to the charging infrastructure. The optimization model will help us determine the optimal location of chargers as well as the number of charging modules at each station over multiple time periods. We define a number of demand dynamics, including flow demand growth varying with charging opportunities on path as well as demand growing naturally with less affect from charging infrastructure, with two objective functions (one maximizing flow coverage and the other maximizing electric vehicle demand). A case study based on a road network around Washington, D.C., New York City, and Boston is presented to provide numerical experiments related to demand dynamics, showing the potential problems in multi-period planning.

38. Md Tanveer Ahmed, Jun Zhuang, Changhyun Kwon (2017). Understanding Conflicting Interests of a Government and a Tobacco Manufacturer: A Game-Theoretic Approach. Group Decision and Negotiation, 26(6), 1209–1230. doi pdf Farming; Subsidy; Food security; Rice; Tobacco; Nash Equilibrium abstract

    Rice is the staple food of nearly half of the population of the world, most of whom live in developing countries. Ensuring a domestic supply of rice from outside sources is difficult for developing countries as less than 5% of the total world’s production is available for international trade. Hence, in order to ensure domestic food security, e.g., food availability and access, governments provide subsidies in agriculture. In many occasions, public money used for the subsidy goes toward promoting undesirable crops like tobacco. Although the strategic interaction between governments and manufacturers is critical, it has not been studied in the literature. This study fills this gap by considering a game between a government (of a developing country) and a tobacco manufacturer in which the government decides on a mix of subsidies and the tobacco manufacturer decides on declaring a purchasing price of tobacco. We provide a numerical study to show that controlling the output harvest price is more effective in reaching the desired end result for both the government and the tobacco manufacturer. A subsidy in fertilizer results in the measurable increase in the government spending but does not have significant effect in reaching the production target. The fertilizer subsidy should be provided only when the output price is too high to be affordable for the population.

39. Masoumeh Taslimi, Rajan Batta, Changhyun Kwon (2017). A Comprehensive Modeling Framework for Hazmat Network Design, Hazmat Response Team Location, and Equity of Risk. Computers & Operations Research, 79, 119–130. doi pdf Hazmat emergency response team; Bi-level network design; Greedy heuristic algorithm; Equity of risk; Robust solution abstract

    This paper considers a bi-level hazmat transportation network design problem in which hazmat shipments have to be transported over a road network between specified origin-destination points. The bi-level framework involves a regulatory authority and hazmat carriers. The control variables for the regulatory authority are locations of hazmat response teams and which additional links to include for hazmat travel. The regulatory authority (upper level) aims to minimize the maximum transport risk incurred by a transportation zone, which is related to risk equity. Our measure of risk incorporates the average response time to the hazmat incidents. Hazmat carriers (lower level) seek to minimize their travel cost. Using optimality conditions, we reformulate the non-linear bi-level model into a single-level mixed integer linear program, which is computationally solvable for medium size problems using a commercial solver. For large size problems, we propose a greedy heuristic approach, which we empirically demonstrate to find good solutions with reasonable computational effort. We also seek a robust solution to capture stochastic characteristics of the model. Experimental results are based on popular test networks from the Sioux Falls and Albany areas.

40. Xiaoping Li, Rajan Batta, Changhyun Kwon (2017). Effective and Equitable Supply of Gasoline to Impacted Areas in the Aftermath of a Natural Disaster. Socio-Economic Planning Sciences, 57, 25–34. doi pdf humanitarian logistics, disaster operations management, location, allocation abstract

    The focus of this research is on supplying gasoline after a natural disaster. There are two aspects for this work: determination of which gas stations should be provided with generators (among those that do not have electric power) and determination of a delivery scheme that accounts for increased demand due to lack of public transportation and considerations such as equity. We develop an MIP for this situation. Two case studies based on Hurricane Sandy in New Jersey are developed and solved in CPLEX.

41. Sung Hoon Chung, Changhyun Kwon (2016). Integrated Supply Chain Management for Perishable Products: Dynamics and Oligopolistic Competition Perspectives with Application to Pharmaceuticals. International Journal of Production Economics, 179, 117–129. doi pdf pharmaceutical supply chain, perishable inventory dynamics, oligopolistic competition, variational inequality, interior point methods abstract

    We propose an integrated supply chain management framework that allows us to explicitly consider the impact of product perishability on a broad scale that includes manufacturers, distribution centers, wholesalers, and demand markets. The framework proposed herein also makes it possible to consider the oligopolistic competition across wholesalers that drives price and demand fluctuations. Furthermore, the supply chain decision rules are derived from necessary conditions in the framework. The inclusion of such salient features allows the framework to generate outcomes that suggest realistic managerial insights. We provide a numerical example in which two multi-national pharmaceutical firms producing a homogeneous medicinal drug and four oligopolistic wholesalers are considered.

42. Longsheng Sun, Mark H. Karwan, Changhyun Kwon (2016). Robust Hazmat Network Design Problems Considering Risk Uncertainty. Transportation Science, 50(4), 1188–1203. doi pdf hazardous materials transportation; network design; robust optimization abstract

    We study robust network design problems for hazardous materials transportation considering risk uncertainty. Risk uncertainty is considered in two ways: (1) uncertainty on each link across all shipments, and (2) uncertainty on each link for each shipment. We extend an existing heuristic framework to solve these two robust network design problems and propose a Lagrangian relaxation heuristic to solve subproblems within the framework. We present our computational experiences and illustrate general insights based on real networks.

43. Iakovos Toumazis, Changhyun Kwon (2016). Worst-case Conditional Value-at-Risk Minimization for Hazardous Materials Transportation. Transportation Science, 50(4), 1174–1187. doi pdf github hazardous materials transportation; conditional value-at-risk; robust optimization abstract

    Despite significant advances in risk management, routing hazardous materials (hazmat) has relied on relatively simpler methods. In this paper, we formally introduce an advanced risk measure, called conditional value-at-risk (CVaR), applied to truck routing problems for hazmat transportation. We find that CVaR offers a flexible, risk-averse, and computationally tractable routing method that is adequate to mitigate hazmat accidents. We further extend CVaR to consider the worst-case CVaR (WCVaR) under data uncertainty. The two important data types in hazmat transportation are accident probabilities and accident consequences, both of which are subject to many ambiguous factors. In addition, historical data are usually insufficient to construct probability distribution of accident probabilities and consequences. This motivates a new robust optimization approach to consider and compute WCVaR. Important axioms are studied for both CVaR and WCVaR risk measures to be coherent and appropriate in the context of hazmat transportation, and computational methods are proposed. We demonstrate the proposed notions of CVaR and WCVaR through a case study in a realistic road network.

44. Anshuman Anjani Kumar, Jee Eun Kang, Changhyun Kwon, Alexander Nikolaev (2016). Inferring Origin-Destination Pairs and Utility-Based Travel Preferences for Shared Mobility Systems Users in a Multi-Modal Environment. Transportation Research Part B: Methodological, 91, 270–291. doi pdf Origin-Destination estimation, Traveler preferences, Expectation Maximization, Probabilistic inference, Multimodal route choice, Bike Sharing Systems, Shared Mobility Systems abstract

    This paper presents a methodological framework to identify population-wide traveler type distribution and simultaneously infer individual travelers’ Origin-Destination (OD) pairs, based on the individual records of a shared mobility (bike) system use in a multimodal travel environment. Given the information about the travelers’ outbound and inbound bike stations under varied price settings, the developed Selective Set Expectation Maximization (SSEM) algorithm infers an underlying distribution of travelers over the given traveler “types” or “classes” treating each traveler’s OD pair as a latent variable; the inferred most likely traveler type for each traveler then informs their most likely OD pair. The experimental results based on simulated data demonstrate high SSEM learning accuracy both on the aggregate and disaggregate levels.

45. Longsheng Sun, Mark H. Karwan, Changhyun Kwon (2016). Implications of Cost Equity Consideration in Hazmat Network Design. Transportation Research Record, 2567(1), 67–77. doi pdf abstract

    The hazmat network design problem (HNDP) aims to reduce the risk of transporting hazmat in the network by enforcing regulation policies. The goal of reducing risk can increase cost for different hazmat carriers. Since HNDP involves multiple parties, it is essential to take the cost increase of all carriers into consideration for the implementation of the regulation policy. While we can consider cost by placing upper bounds on the total increase, the actual cost increase for various OD pairs can differ, which results in unfairness among carriers. Thus we propose to consider the cost equity issue as well in HNDP. Additionally, due to the existence of multiple solutions in current HNDP models and the possibility of unnecessarily closing road segments, we introduce a new objective considering the length of all the closed links. Our computational experience is based on a real network and we show results under different cost consideration cases.

46. Tolou Esfandeh, Changhyun Kwon, Rajan Batta (2016). Regulating Hazardous Materials Transportation by Dual Toll Pricing. Transportation Research Part B: Methodological, 83, 20–35. doi pdf hazardous material transportation; toll setting; non-convex optimization; bi-level programming abstract

    We investigate dual-toll setting as a policy tool to mitigate the risk of hazardous material (hazmat) shipment in road networks. We formulate the dual-toll problem as a bi-level program wherein the upper level aims at minimizing the risk, and the lower level explores the user equilibrium decision of the regular vehicles and hazmat carriers given the toll. When the upper level objective is to minimize the risk and all links are tollable, we decompose the formulation into first-stage and second-stage, and suggest a computational method to solve each stage. Our two-stage solution methodology guarantees nonnegative valid dual tolls regardless of the solution accuracy of the first-stage problem. We also consider a general dual-toll setting problem where the regulator rather wishes to minimize a combination of risk and the paid tolls and/or some links are untollable. To solve this truly bilevel problem, we provide heuristic algorithms that decompose the problem into subproblems each being solved by a line search. Case studies based on the Sioux Falls network illustrate the insights on the dual-toll policies.

47. Longsheng Sun, Mark H. Karwan, Changhyun Kwon (2016). Incorporating Driver Behaviors in Network Design Problems: Challenges and Opportunities. Transport Reviews, 36(4), 454–478. doi pdf network design; behavior route choice; random utility; random regret; bounded rationality; cumulative prospect theory; fuzzy logic; dynamic learning; SILK theory abstract

    The goal of network design problem (NDP) is to make optimal decisions to achieve a certain objective such as minimizing total travel time or maximizing tolls collected in the network. A critical component to NDP is how travelers make their route choices. Researchers in transportation have adopted human decision theories to describe more accurate route choice behaviors. In this paper, we review the NDP with various route choice models: the random utility model (RUM), Random Regret-Minimization (RRM) model, bounded rationality (BR), cumulative prospect theory (CPT), the fuzzy logic model (FLM) and dynamic learning models (DLM). Moreover, we identify challenges in applying behavioral route choice models to NDP and opportunities for future research.

48. Ha Hwang, JiYoung Park, Changhyun Kwon, Kathryn Friedman, Nathan Attard, Shen Hao Chang, Samuel Wells (2015). The Ties that Bind: Economic and Freight Transportation Implications of U.S.-Canada Border Bridges Using a Bi-national Transportation Network-Combined Economic Model. Research in Transportation Business & Management, 16, 32–49. doi pdf Cross-border freight; trade network analysis; border wait time; Bi-national TransNIEMO abstract

    This study combines US-Canada bi-national highway network data with a freight flow dataset using ports of entry (POE) via highway border crossings. Through several sub-procedures, the US and Canada highway systems are integrated into a single network dataset. In addition, border wait time dataset was monitored and analyzed to set the border delay baseline. This dataset enables us to explore the freight traffic pattern between the US and Canada. Weighted Eigenvector Score is computed using a Social Network Analysis tool. The results demonstrate that major regional bodies are the primary users of major POE between the US and Canada. This study not only offers an improved understanding of the economic implications of US-Canada border crossings, but also contributes to developing a simulation tool, a bi-national Transportation-combined National Interstate Economic Model. Such a tool is expected to extend and apply to other contexts, such as transportation and national and bi-national security, among other applications. Additionally, this study suggests several important considerations for US and Canadian officials charged with devising policy to protect against security threats while facilitating legitimate flows of goods, services and people across the border.

49. Vijay Ganesh Hariharan, Debabrata Talukdar, Changhyun Kwon (2015). Optimal Targeting of Advertisement for New Products with Multiple Consumer Segments. International Journal of Research in Marketing, 32(3), 263–271. doi pdf ssrn new product diffusion; advertisement; targeting; social contagion; dynamic optimization abstract

    Armed with improved targeting technology, firms are increasingly interested to optimize their advertising dollars through consumer segment-specific targeting, particularly while introducing new products. That task becomes especially important in markets with distinct consumer segments – the early market and the main market – that affect each other’s adoption behavior. In this study, in contrast to prior normative studies that assume a single-segment market structure, we derive dynamic optimal advertising and segment-specific targeting strategies for firms facing a two-segment market structure. We allow for mutual demand interactions between the two segments, and for the diffusion parameters, advertising sensitivity, and cost of targeting to differ across the segments. We model the effect of advertising as a logarithmic function that accounts for diminishing marginal returns. Among our key findings: From profit optimization perspective, our two-segment model outperforms the single-segment model under multiple diffusion dynamics contexts – especially for the ‘bimodal chasm’ and the ‘early dip followed by bell-shaped’ type diffusion patterns – even when the cost of targeting the early market is relatively high. Our numerical analyses indicate that the optimal share of advertisement targeted to the early market segment at launch needs to be much higher than the share of the early market segment in the population. Advertising sensitivity, relative cost of targeting the early market, and the proportion of early market consumers in the population have the greatest effects on the optimal time to transition the targeted advertising spending from the early to the main market segment.

50. Sung Hoon Chung, Changhyun Kwon (2015). Multi-Period Planning for Electric-Car Charging Station Locations: a Case of Korean Expressways. European Journal of Operational Research, 242(2), 677–687. doi pdf github flow-refueling location; electric vehicles; multi-period planning; Korean Expressways abstract

    One of the most critical barriers to widespread adoption of electric cars is the lack of charging station infrastructure. Although it is expected that a sufficient number of charging stations will be constructed eventually, due to various practical reasons they may have to be introduced gradually over time. In this paper, we formulate a multi-period optimization model based on a flow-refueling location model for strategic charging station location planning. We also propose two myopic methods and develop a case study based on the real traffic flow data of the Korean Expressway network in 2011. We discuss the performance of the three proposed methods.

51. Taehan Lee, Changhyun Kwon (2014). A Short Note on the Robust Combinatorial Optimization Problems with Cardinality Constrained Uncertainty. 4OR, 12(4), 373–378. doi pdf github robust combinatorial optimization; discrete optimization abstract

    Robust combinatorial optimization problems with cardinality constrained uncertainty may be solved by a finite number of nominal problems. In this paper, we show that the number of nominal problems to be solved can be reduced significantly.

52. JiYoung Park, Changhyun Kwon, Minsu Son (2014). Economic Implications of the Canada-U.S. Border Bridges: Applying a Binational Economic Model for Canada and the U.S.. Research in Transportation Business and Management, 11, 123–133. doi pdf Border bridges; congestion; freight transportation; economic costs abstract

    This study provides an approach to measure the economic costs stemming from delays on the bridges connecting the U.S. and Canada. We selected the busiest bridges in the U.S. and Canada that connects the Buffalo-Niagara Metropolitan region and the Ontario province. Separated by the Great Lakes and waterways, Ontario has a significant portion of its trade activities with the U.S. by way of freight transportation crossing border bridges connecting the two countries. Using binational economic models, we found that the economic implications of the Canada-U.S. border bridges are in the range of 120,000to400,000 per day in total. Furthermore, the binational economic models we developed have provided which industries are most impacted from the freight delays on the bridges based on diverse scenarios. Our modeling approach and scenario development process provide diverse simulation tests with the changes of freight transportation costs and patterns for key sectors.

53. Yingying Kang, Rajan Batta, Changhyun Kwon (2014). Value-at-Risk Model for Hazardous Material Transportation. Annals of Operations Research, 222(1), 361–387. doi pdf github hazardous materials transportation; value-at-risk; social risk mitigation abstract

    This paper introduces a Value-at-Risk (VaR) model to support route choices in hazmat transportation under a specified risk confidence level. It establishes several properties of the model and proposes an exact solution procedure for single trips. Through a case study in the Albany, NY region, different routes are generated depending on the confidence level, highlighting how risk tolerance shapes route selection.

54. Paul G. Berglund, Changhyun Kwon (2014). Solving a Location Problem of a Stackelberg Firm Competing with Cournot-Nash Firms. Networks and Spatial Economics, 14(1), 117–132. doi pdf Location Analysis; Stackelberg-Cournot-Nash Equilibrium; Game Theory; Variational Inequality; Simulated Annealing abstract

    We study a discrete facility location problem on a network, where the locating firm acts as the leader and other competitors as the followers in a Stackelberg-Cournot-Nash game. To maximize expected profits the locating firm must solve a mixed-integer problem with equilibrium constraints. Finding an optimal solution is hard for large problems, and full-enumeration approaches have been proposed in the literature for similar problem instances.

55. Yingying Kang, Rajan Batta, Changhyun Kwon (2014). Generalized Route Planning Model for Hazardous Material Transportation with VaR and Equity Considerations. Computers & Operations Research, 43, 237–247. doi pdf github Value-at-Risk; multi-trip hazmat transportation; social risk mitigation; risk equity; dissimilar path abstract

    Recently, the Value-at-Risk (VaR) framework was introduced for the routing problem of a single hazmat trip. In this paper, we extend the VaR framework in two important ways. First, we show how to apply the VaR concept to a more realistic multi-trip multi-hazmat type framework, which determines routes that minimize the global VaR value while satisfying equity constraints. Second, we show how to embed the algorithm for the single hazmat trip problem into a Lagrangian relaxation framework to obtain an efficient solution method for this general case. We test our computational experience based on a real-life hazmat routing scenario in the Albany district of New York State. Our results indicate that one can achieve a high degree of risk dispersion while controlling the VaR value within the desired confidence level.

56. Paul G. Berglund, Changhyun Kwon (2014). Robust Facility Location Problem for Hazardous Waste Transportation. Networks and Spatial Economics, 14(1), 91–116. doi pdf Location problem; Hazardous waste facility; Robust optimization; Genetic algorithm abstract

    We consider a robust facility location problem for hazardous materials (hazmat) transportation considering routing decisions of hazmat carriers. Given a network and a known set of nodes from which hazmat originate, we compute the locations of hazmat processing sites (e.g. incinerators) which will minimize total cost, in terms of fixed facility cost, transportation cost, and exposure risk. We assume that hazmat will be taken to the closest existing processing site. We present an exact full enumeration method, which is useful for small or medium-size problems. For larger problems, the use of a genetic algorithm is explored. Through numerical experiments, we discuss the impact of uncertainty and robust optimization in the hazmat combined location-routing problem.

57. Md Tanveer Ahmed, Changhyun Kwon (2014). Optimal Contract-Sizing in Online Display Advertising for Publishers with Regret Considerations. Omega, 42(1), 201–212. doi pdf Newsboy problem; Operations management; Risk abstract

    In this paper, we study optimal contract problems for online display advertisements with pay-per-view pricing schemes. We first provide and analyze a single-contract model equivalent to the newsvendor problem. We then consider a stochastic optimization with two different advertisements and show how regret considerations of a risk-averse publisher can lead to different optimal contract-sizing strategies. Numerical experiments illustrate changes in pricing and contract allocation based on publisher risk preferences.

58. Iakovos Toumazis, Changhyun Kwon (2013). Routing Hazardous Materials on Time-Dependent Networks using Conditional Value-at-Risk. Transportation Research Part C: Emerging Technologies, 37, 73–92. doi pdf github dynamic shortest path; time-dependent network; conditional value-at-risk; hazardous materials transportation abstract

    We propose a new method for mitigating risk in routing hazardous materials (hazmat), based on the conditional value-at-risk (CVaR) measure on time-dependent vehicular networks. The CVaR models are shown to be flexible and general routing models for hazmat transportation, and can be solved efficiently. This paper extends the previous research by considering CVaR for hazmat transportation in the case where accident probabilities and accident consequences are time-dependent. We provide a numerical method to determine an optimal departure time and an optimal route for a given origin-destination pair. The proposed algorithm is tested in a realistic road network in Buffalo, NY, USA and the results are discussed.

59. Changhyun Kwon, Taehan Lee, Paul G. Berglund (2013). Robust Shortest Path Problems with Two Uncertain Multiplicative Cost Coefficients. Naval Research Logistics, 60(5), 375–394. doi pdf github robust shortest path; budgeted uncertainty; hazardous materials transportation abstract

    We consider a robust shortest path problem when the cost coefficient is the product of two uncertain factors. We first show that the robust problem can be solved in polynomial time by a dual variable enumeration with shortest path problems as subproblems. We also propose a path enumeration approach using a K-shortest paths finding algorithm that may be efficient in many real cases. An application in hazardous materials transportation is discussed and the solution methods are illustrated by numerical examples.

60. Md Tanveer Ahmed, Changhyun Kwon (2012). Pricing Game of Online Display Advertisement Publishers. European Journal of Operational Research, 219, 477–487. doi pdf online display advertising; optimal pricing; Nash equilibrium; newsvendor problem abstract

    We consider an online display advertisement publisher who maximizes the revenue by optimal pricing in an oligopoly setting. Each publisher interacts with others though setting cost-perimpression (CPM) that affects the demand for everyone. Using the pseudoconcavity of the objective function, we study the best response of the publisher while her strategy space changes. We also consider the sensitivity of the publisher while other publishers changes their CPM. In both cases, the best response of the publisher depends entirely on her current best response CPM. We provide an algorithm for finding the equilibrium and illustrate by numerical examples.

61. Anand Srinivasan, Changhyun Kwon (2012). Operations of Online Advertising Services and Publisher's Option. Journal of the Operational Research Society, 63, 674–682. doi pdf online advertising; Nash bargaining game; option contract abstract

    We analyse the use of options for online advertisement publishers. By providing a discount or rewards to advertisers, publishers can utilize their uncertain service capacity, page-views, more efficiently. We use Generalized Nash Bargaining to study the feasibility of the option contract and solve for an optimal value for the option price. We compare the revenues and benefits from advertisements under the option contract, with those without the options using numerical studies. We also study the impact of pricing and other components in the game on the optimal option price, the publisher’s revenues, and the advertiser’s benefits from the advertisements.

62. Byung Do Chung, Jiahan Li, Tao Yao, Changhyun Kwon, Terry L. Friesz (2012). Demand Learning and Dynamic Pricing under Competition in a State-Space Framework. IEEE Transactions on Engineering Management, 59(2), 240–249. doi pdf Competition, demand learning, differential variational inequality, dynamic pricing, Markov chain Monte Carlo, nonlinear time series abstract

    In this paper, we propose a revenue optimization framework integrating demand learning and dynamic pricing for firms in monopoly or oligopoly markets. We introduce a state-space model for this revenue management problem, which incorporates game-theoretic demand dynamics and nonparametric techniques for estimating the evolution of underlying state variables. Under this framework, stringent model assumptions are removed. We develop a new demand learning algorithm using Markov chain Monte Carlo methods to estimate model parameters, unobserved state variables, and functional coefficients in the nonparametric part. Based on these estimates, future price sensitivities can be predicted, and the optimal pricing policy for the next planning period is obtained. To test the performance of demand learning strategies, we solve a monopoly firm’s revenue maximizing problem in simulation studies. We then extend this paradigm to dynamic competition, where the problem is formulated as a differential variational inequality. Numerical examples show that our demand learning algorithm is efficient and robust.

63. Jiashan Wang, Yingying Kang, Changhyun Kwon, Rajan Batta (2012). Dual Toll Pricing for Hazardous Material Transport with Linear Delay. Networks and Spatial Economics, 12(1), 147–165. doi pdf hazardous materials; toll pricing; congestion; risk abstract

    In this paper, we propose a dual toll pricing method to mitigate risk of hazardous materials (hazmat) transportation. We aim to simultaneously control both regular and hazmat vehicles to reduce the risk. In our model, we incorporate a new risk measure to consider duration-population-frequency of hazmat exposure. We first formulate the model as a Mathematical Program with Equilibrium Constraints (MPEC). Then we decompose the MPEC formulation into first-stage and second-stage problems. Separate methods are developed to solve each stage. A numerical example is provided and possible extensions are discussed.

64. Taewoo Jung, Changhyun Kwon (2011). Retailer-Supplier Matching: an Application of the Deferred Acceptance Algorithm. International Journal of Services Operations and Informatics, 6(3), 248–258. doi pdf abstract

    In this paper, we apply matching theory to supply chain coordination. We present mathematical optimization models similar to the newsvendor problem to provide appropriate conditions for retailer-supplier matching. In particular, our matching algorithm, compared to the general matching theory, has uniquely been affected by contract sizes and ordering sequences. We also study that our matching application guarantees stable and optimal outcomes. Numerical examples with various parameter settings are provided to test the feasibility of the matching algorithms. We find that we can avoid the worst matching case when we use the proposed matching algorithms.

65. Terry L. Friesz, Taeil Kim, Changhyun Kwon, Matthew A. Rigdon (2011). Approximate Network Loading and Dual Time Scale Dynamic User Equilibrium. Transportation Research Part B: Methodological, 45(1), 176–207. doi pdf dynamic user equilibrium; differential variational inequalities; differential algebraic equations; dual-time-scale; fixed-point algorithm in Hilbert space abstract

    In this paper we present a dual-time-scale formulation of dynamic user equilibrium (DUE) with demand evolution. Our formulation belongs to the problem class that Pang and Stewart (2008) refer to as differential variational inequalities. It combines the within-day time scale for which route and departure time choices fluctuate in continuous time with the day-to-day time scale for which demand evolves in discrete time steps. Our formulation is consistent with the often told story that drivers adjust their travel demands at the end of every day based on their congestion experience during one or more previous days. We show that analysis of the within-day assignment model is tremendously simplified by expressing dynamic user equilibrium as a differential variational inequality. We also show there is a class of day-to-day demand growth models that allow the dual-time-scale formulation to be decomposed by time-stepping to yield a sequence of continuous time, single-day, dynamic user equilibrium problems. To solve the single-day DUE problems arising during time-stepping, it is necessary to repeatedly solve a dynamic network loading problem. We observe that the network loading phase of DUE computation generally constitutes a differential algebraic equation (DAE) system, and we show that the DAE system for network loading based on the link delay model (LDM) of Friesz et al. (1993) may be approximated by a system of ordinary differential equations (ODEs). That system of ODEs, as we demonstrate, may be efficiently solved using traditional numerical methods for such problems. To compute an actual dynamic user equilibrium, we introduce a continuous time fixed-point algorithm and prove its convergence for effective path delay operators that allow a limited type of nonmonotone path delay. We show that our DUE algorithm is compatible with network loading based on the LDM and the cell transmission model (CTM) due to Daganzo (1995). We provide a numerical example based on the much studied Sioux Falls network.

66. Yongma Moon, Changhyun Kwon (2011). Online Advertisement Service Pricing and an Option Contract. Electronic Commerce Research and Applications, 10(1), 38–48. doi pdf online advertisements, option contract, Nash bargaining, click-through rate, utility maximization abstract

    For the Internet advertisement market, we consider a contract problem between advertisers and publishers. Among several ways of pricing online advertisements, the methods based on cost-per-impression (CPM) and cost-per-click (CPC) are the two most popular. The CPC fee is proportional to the click-through rate (CTR), which is uncertain and makes decisions of advertisers and publishers difficult. In this paper, we suggest a hybrid pricing scheme: advertisers pay the minimum of CPM and CPC fees by purchasing an option from publishers. To determine the option price, we consider a Nash bargaining game for negotiation between an advertiser and a publisher and provide the solution. Further, we show that such option contracts will help the advertiser avoid high cost and the publisher generate more revenue. The option contract will also improve the contract feasibility, compared to CPM and CPC.

67. Changhyun Kwon (2011). Single-Period Balancing of Pay-Per-Click and Pay-Per-View Online Display Advertisements. Journal of Revenue and Pricing Management, 10(3), 261–270. doi pdf online advertising; display advertisements; cost-per-impression; cost-per-click; click-through-rate; web publisher abstract

    In this article, we study a balancing problem of web publishers for pay-per-view and pay-per-click contracts for online display advertising. Considering the details of contracts, we refine prior research results on the recommendation of optimal strategies. We examine the problem by formalizing a simple stochastic optimization problem for a single period of advertising contracts. We investigate how pricing and other contract components will affect the optimal display strategies analytically and numerically.

68. Changhyun Kwon, Terry L. Friesz, Reetabrata Mookherjee, Tao Yao, Baichun Feng (2009). Non-cooperative Competition Among Revenue Maximizing Service Providers with Demand Learning. European Journal of Operational Research, 197(3), 981–996. doi pdf Revenue management, Pricing, Demand learning, Differential games, Kalman filters abstract

    This paper recognizes that in many decision environments in which revenue optimization is attempted, an actual demand curve and its parameters are generally unobservable. Herein, we describe the dynamics of demand as a continuous time differential equation based on an evolutionary game theory perspective. We then observe realized sales data to obtain estimates of parameters that govern the evolution of demand; these are refined on a discrete time scale. The resulting model takes the form of a differential variational inequality. We present an algorithm based on a gap function for the differential variational inequality and report its numerical performance for an example revenue optimization problem.

69. Changhyun Kwon, Terry L. Friesz (2008). Valuation of American Options by the Gradient Projection Method. Applied Mathematics and Computation, 206(1), 380–388. doi pdf American options, variational inequalities, gradient projection abstract

    We study an equivalent optimization problem with an inequality constraint and boundary conditions, whose necessary condition for optimality is the variational inequality presentation of American options. To solve the problem, we use the gradient projection method, with discretizations both in time and space. We tested the algorithm and compared with the projective successive over-relaxation method.

70. Terry L. Friesz, Changhyun Kwon (2008). Supply Chain Design in Perfect Competition. International Journal of Services Operations and Informatics, 3(3/4), 340–356. doi pdf abstract

    In this paper, we apply the theory of optimal control and theory of traffic assignment for supply chain design in perfect competition. We model the time staging and generalised routing of input factors needed for production by a firm. The production process will typically involve several stages, and as such is described by paths through a production network whose nodes are the various stages of production. We develop an algorithm and test it for a small numerical example.

71. Terry L. Friesz, Changhyun Kwon, Reetabrata Mookherjee (2007). A Computable Theory of Dynamic Congestion Pricing. In: Transportation and Traffic Theory 2007: Papers selected for presentation at ISTTT17, a peer reviewed series since 1959, R. E. Allsop, M. G. H. Bell and B. G. Heydecker (Eds.), pp. 1-26. pdf abstract

    In this paper we present a theory of dynamic congestion pricing for the day-to-day as well as the within-day time scales. The equilibrium design problem emphasized herein takes the form of an MPEC, which we call the Dynamic Optimal Toll Problem with Equilibrium Constraints, or DOTPEC. The DOPTEC formulation we employ recalls an important earlier result that allows the equilibrium design problem to be stated as a single level problem, a result which is surprisingly little known. The DOPTEC maintains the usual design objective of minimizing the system travel cost by appropriate toll pricing. We describe how an infinite dimensional mathematical programming perspective may be employed to create an algorithm for the DOTPEC. A numerical example is provided.

72. Tan C. Miller, Terry L. Friesz, Roger L. Tobin, Changhyun Kwon (2007). Reaction Function Based Dynamic Location Modeling in Stackelberg-Nash-Cournot Competition. Networks and Spatial Economics, 7(1), 77–97. doi pdf Dynamic Stackelberg equilibrium location modeling, reaction functions abstract

    We formulate a dynamic facility location model for a firm locating on a discrete network. It is assumed that this locating firm will act as the leader firm in an industry characterized by Stackelberg leader-follower competition. The firm’s I competitors are assumed to act as Cournot firms and are each assumed to operate under the assumption of zero conjectural variation with respect to their I-1 Cournot competitors. Using sensitivity analysis of variational inequalities within a hierachical mathematical programming approach, we develop reaction function based dynamic models to optimize the Stackelberg firm’s location decision. In the second half of this paper, we use these models to illustrate through a numerical example the insights yielded by our approach.

Papers in Refereed Proceedings

1. Jiwoo Son, Zhikai Zhao, Federico Berto, Chuanbo Hua, Zhiguang Cao, Changhyun Kwon, Jinkyoo Park. RRNCO: Towards Real-World Routing with Neural Combinatorial Optimization The Fourteenth International Conference on Learning Representations, ICLR 2026 url abstract

   The practical deployment of Neural Combinatorial Optimization (NCO) for Vehicle Routing Problems (VRPs) is hindered by a critical sim-to-real gap. This gap stems not only from training on oversimplified Euclidean data but also from node-based architectures incapable of handling the node-and-edge-based features with correlated asymmetric cost matrices, such as those for real-world distance and duration. We introduce RRNCO, a novel architecture specifically designed to address these complexities. RRNCO’s novelty lies in two key innovations. First, its Adaptive Node Embedding (ANE) efficiently fuses spatial coordinates with real-world distance features using a learned contextual gating mechanism. Second, its Neural Adaptive Bias (NAB) is the first mechanism to jointly model asymmetric distance, duration, and directional angles, enabling it to capture complex, realistic routing constraints. Moreover, we introduce a new VRP benchmark grounded in real-world data crucial for bridging this sim-to-real gap, featuring asymmetric distance and duration matrices from 100 diverse cities, enabling the training and validation of NCO solvers on tasks that are more representative of practical settings. Experiments demonstrate that RRNCO achieves state-of-the-art performance on this benchmark, significantly advancing the practical applicability of neural solvers for real-world logistics.

2. Chuanbo Hua, Federico Berto, Zhikai Zhao, Jiwoo Son, Changhyun Kwon, Jinkyoo Park. USPR: Learning a Unified Solver for Profiled Routing. Proceedings of the AAAI Conference on Artificial Intelligence, 40, AAAI 2026. arXiv abstract

   The Profiled Vehicle Routing Problem (PVRP) extends the classical VRP by incorporating vehicle-client-specific preferences and constraints, reflecting real-world requirements such as zone restrictions and service-level preferences. While recent reinforcement-learning solvers have shown promising performance, they require retraining for each new profile distribution, suffer from poor representation ability, and struggle to generalize to out-of-distribution instances. In this paper, we address these limitations by introducing Unified Solver for Profiled Routing (USPR), a novel framework that natively handles arbitrary profile types. USPR introduces on three key innovations: (i) Profile Embeddings (PE) to encode any combination of profile types; (ii) Multi-Head Profiled Attention (MHPA), an attention mechanism that models rich interactions between vehicles and clients; (iii) Profile-aware Score Reshaping (PSR), which dynamically adjusts decoder logits using profile scores to improve generalization. Empirical results on diverse PVRP benchmarks demonstrate that USPR achieves state-of-the-art results among learning-based methods while offering significant gains in flexibility and computational efficiency. We make our source code publicly available to foster future research.

3. Sohyun Jeong, Changhyun Kwon, Hyun-Jung Kim. Self-Supervised Job Grouping for Parallel Machine Scheduling with Family Setups Proceedings of the 11th IFAC Conference on Manufacturing Modelling, Management and Control MIM 2025. doi abstract

   This study addresses the parallel machine scheduling problem with family setups in large-scale manufacturing, motivated by challenges in the air conditioner assembly process, aiming to minimize total tardiness and setup costs. The proposed framework integrates self-supervised learning for job grouping and incorporates tardiness evaluator for iterative refinement. Experimental evaluations on datasets inspired by real-world scenarios demonstrate that the proposed framework effectively reduces total tardiness and setup costs compared to baseline methods. These results highlight the framework’s scalability, computational efficiency, and potential as a practical solution to real-world scheduling challenges.

4. Federico Berto, Chuanbo Hua, Laurin Luttmann, Jiwoo Son, Junyoung Park, Kyuree Ahn, Changhyun Kwon, Lin Xie, Jinkyoo Park. PARCO: Learning parallel autoregressive policies for efficient multi-agent combinatorial optimization. Advances in Neural Information Processing Systems 39: Annual Conference on Neural Information Processing Systems 2025, NeurIPS 2025. arXiv github abstract

   Multi-agent combinatorial optimization problems such as routing and scheduling have great practical relevance but present challenges due to their NP-hard combinatorial nature, hard constraints on the number of possible agents, and hard-to-optimize objective functions. This paper introduces PARCO (Parallel AutoRegressive Combinatorial Optimization), a novel approach that learns fast surrogate solvers for multi-agent combinatorial problems with reinforcement learning by employing parallel autoregressive decoding. We propose a model with a Multiple Pointer Mechanism to efficiently decode multiple decisions simultaneously by different agents, enhanced by a Priority-based Conflict Handling scheme. Moreover, we design specialized Communication Layers that enable effective agent collaboration, thus enriching decision-making. We evaluate PARCO in representative multi-agent combinatorial problems in routing and scheduling and demonstrate that our learned solvers offer competitive results against both classical and neural baselines in terms of both solution quality and speed. We make our code openly available at https://github.com/ai4co/parco.

5. Federico Berto, Chuanbo Hua, Junyoung Park, Laurin Luttmann, Yining Ma, Fanchen Bu, Jiarui Wang, Haoran Ye, Minsu Kim, Sanghyeok Choi, Nayeli Gast Zepeda, Andre Hottung, Jianan Zhou, Jieyi Bi, Yu Hu, Fei Liu, Hyeonah Kim, Jiwoo Son, Haeyeon Kim, Davide Angioni, Wouter Kool, Zhiguang Cao, Qingfu Zhang, Joungho Kim, Jie Zhang, Kijung Shin, Cathy Wu, Sungsoo Ahn, Guojie Song, Changhyun Kwon, Kevin Tierney, Lin Xie, Jinkyoo Park (2025) RL4CO: an Extensive Reinforcement Learning for Combinatorial Optimization Benchmark. KDD ’25: Proceedings of the 31st ACM SIGKDD Conference on Knowledge Discovery and Data Mining, V.2, 5278–5289. arXiv doi github abstract

   Deep reinforcement learning (RL) has recently shown significant benefits in solving combinatorial optimization (CO) problems, reducing reliance on domain expertise, and improving computational efficiency. However, the field lacks a unified benchmark for easy development and standardized comparison of algorithms across diverse CO problems. To fill this gap, we introduce RL4CO, a unified and extensive benchmark with in-depth library coverage of 23 state-of-the-art methods and more than 20 CO problems. Built on efficient software libraries and best practices in implementation, RL4CO features modularized implementation and flexible configuration of diverse RL algorithms, neural network architectures, inference techniques, and environments. RL4CO allows researchers to seamlessly navigate existing successes and develop their unique designs, facilitating the entire research process by decoupling science from heavy engineering. We also provide extensive benchmark studies to inspire new insights and future work. RL4CO has attracted numerous researchers in the community and is open-sourced at https://github.com/ai4co/rl4co.

6. Hyeonah Kim, Sanghyeok Choi, Jiwoo Son, Jinkyoo Park, Changhyun Kwon (2025) Neural Genetic Search in Discrete Spaces. ICML ’25: Proceedings of the 42nd International Conference on Machine Learning, in Proceedings of Machine Learning Research, 267:29969–29987. arXiv url github abstract

   Effective search methods are crucial for improving the performance of deep generative models at test time. In this paper, we introduce a novel test-time search method, Neural Genetic Search (NGS), which incorporates the evolutionary mechanism of genetic algorithms into the generation procedure of deep models. The core idea behind NGS is its crossover, which is defined as parent-conditioned generation using trained generative models. This approach offers a versatile and easy-to-implement search algorithm for deep generative models. We demonstrate the effectiveness and flexibility of NGS through experiments across three distinct domains: routing problems, adversarial prompt generation for language models, and molecular design.

7. Chuanbo Hua, Federico Berto, Jiwoo Son, Seunghyun Kang, Changhyun Kwon, Jinkyoo Park (2025) CAMP: Collaborative Attention Model with Profiles for Vehicle Routing Problems. AAMAS ’25: Proceedings of the 2025 International Conference on Autonomous Agents and Multiagent Systems, 1015–1024 arXiv github abstract

   The profiled vehicle routing problem (PVRP) is a generalization of the heterogeneous capacitated vehicle routing problem (HCVRP) in which the objective is to optimize the routes of vehicles to serve client demands subject to different vehicle profiles, with each having a preference or constraint on a per-client basis. While existing learning methods have shown promise for solving the HCVRP in real time, no learning method exists to solve the more practical and challenging PVRP. In this paper, we propose a Collaborative Attention Model with Profiles (CAMP), a novel approach that learns efficient solvers for PVRP using multi-agent reinforcement learning. CAMP employs a specialized attention-based encoder architecture to embed profiled client embeddings in parallel for each vehicle profile. We design a communication layer between agents for collaborative decision-making across profiled embeddings at each decoding step and a batched pointer mechanism to attend to the profiled embeddings to evaluate the likelihood of the next actions. We evaluate CAMP on two variants of PVRPs: PVRP with preferences (PRVP-P), which explicitly influence the reward function, and PVRP with zone constraints (PVRP-ZC) scenarios with different numbers of agents and nodes, demonstrating that our learned solvers achieve competitive results compared to both classical state-of-the-art neural multi-agent models in terms of solution quality and computational efficiency. We make our code openly available at https://github.com/ai4co/camp.

8. Sobhanan, A., S. Mahmoudinazlou, H. Charkhgard, C. Kwon. (2024) A branch-and-price algorithm for emergency humanitarian logistics with a mixed truck-drone fleet. Proceedings of the IISE Annual Conference & Expo 2024, 6pages. (Finalist for the 2024 IISE OR Track Best Paper Award) pdf abstract

   Humanitarian aid distribution often prioritizes rapid relief operations or emergency services under time constraints, as opposed to commercial transportation problems, where the primary objective is to minimize operational costs. Drones can offer immense potential to achieve this goal by leveraging their aerial mobility. Specifically, drones can surpass ground transportation and navigate directly through disrupted or inaccessible roads, ensuring the quickest path to deliver aid where the ground vehicle may face obstacles. However, drones have limitations in terms of flying range and load capacity. To effectively provide time-sensitive emergency services, combining a ground vehicle with one or more aerial vehicles enhances coverage. Our approach integrates a truck as a mobile depot for multiple drones, where a drone battery is replenished on landing after a flight, and the fleet operates in tandem to serve the locations visited. We formulate a mixed-integer linear programming (MILP) model to maximize the weighted sum of locations served by this mixed truck-drone fleet under time constraints. We further develop a branch-and-price algorithm to solve this problem, where the pricing subproblem is solved using dynamic programming recursions with dominance rules. Our results demonstrate the computational superiority of this method compared to a commercial optimization solver and its potential for expediting aid distribution during an emergency.

9. Junyoung Park, Changhyun Kwon, Jinkyoo Park (2023) Learn to solve the min-max multiple traveling salesmen problem with reinforcement learning. Proceedings of the 2023 International Conference on Autonomous Agents and Multiagent Systems (AAMAS 2023), 878—886, 9 pages. url pdf abstract

   We propose ScheduleNet, a scalable scheduler that minimizes the task completion time by coordinating multiple agents. We formulate the min-max multiple traveling salesmen problem (mTSP) as a Markov decision process with an episodic reward and derive a scalable decision-making policy using reinforcement learning (RL). The decision-making procedure of ScheduleNet includes: (1) representing the state of a problem with the agent-task graph, (2) extracting node embedding for agents and tasks by employing the type-aware graph attention, (3) and computing the task assignment probability with the computed node embedding. We show that ScheduleNet can outperform other heuristic approaches and existing deep RL approaches, particularly validating its exceptional effectiveness in solving large and practical problems. We also confirm that ScheduleNet can effectively solve practical mTSP variants, which include limited observation and online mTSP.

10. C. Kwon. (2011) Conditional Value-at-Risk Model for Hazardous Materials Transportation, in Proceedings of the 2011 Winter Simulation Conference, S. Jain, R. R. Creasey, J. Himmelspach, K. P. White, and M. Fu, eds. pp. 1708-1714 doi abstract

    This paper investigates how the conditional value-at-risk (CVaR) can be used to mitigate risk in hazardous materials (hazmat) transportation. Routing hazmat must consider accident probabilities and accident consequences that depend on the hazmat types and route choices. This paper proposes a new method for mitigating risk based on CVaR measure. While the CVaR model is popularly used in financial portfolio optimization problems, its application in hazmat transportation is new. A computational method for determining the optimal CVaR route is proposed and illustrated by a case study in the road network surrounding Albany, NY.

11. Jeong, B., C. Kwon, Y.S. Zhang, and S. Yang. (2010) Evaluating the Research Performance of a R\&D Program: an Application of DEA. Proceedings of the International Multi-Conference of Engineers and Computer Scientists 2010 Vol III, 5 pages. pdf

12. Friesz, T. L., M. A. Rigdon and C. Kwon. (2008) A New Gradient Projection Algorithm for the Dynamic User Equilibrium Problem. International Symposium on Dynamic Traffic Assignment 2008, 9 pages.

13. Friesz, T. L., C. Kwon and M. A. Rigdon. (2007) A Model of Competition in Service Management. Proceedings of 2007 IEEE/INFORMS International Conference on Service Operations and Logistics, and Informatics, 6 pages. doi

14. Friesz, T. L., C. Kwon, A. Chow and B. Heydecker. (2007) A Dynamic Efficient Toll Problem. The Sixth Triennial Symposium on Transportation Analysis (TRISTAN VI), 4 pages pdf

15. Friesz, T. L., R. Mookherjee and C. Kwon. (2006) Dynamic Non-cooperative Games as a Foundation for Modeling Dynamic User Equilibrium. The First International Symposium on Dynamic Traffic Assignment, 18 pages. pdf

16. Friesz, T. L., C. Kwon and R. Mookherjee. (2006) A Computable Theory of Dynamic Congestion Pricing The First International Symposium on Dynamic Traffic Assignment, 15 pages. pdf

Book Chapters

1. Chung, S. H. and C. Kwon (2022), Drone-Truck Routing Problems, in Encyclopedia of Optimization (Eds.:P. Pardalos and O. Prokopyev), Springer pdf

2. Toumazis, I.^*, C. Kwon, and R. Batta (2013), Value-at-Risk and Conditional Value-at-Risk Minimization for Hazardous Materials Routing, in Handbook of OR/MS Models in Hazardous Materials Transportation (Eds.:R. Batta and C. Kwon), Springer doi pdf github

3. Friesz, T. L., T. I. Kim, C. Kwon and M. A. Rigdon (2010), Computing Dual Time Scale Dynamic User Equilibria, in New Developments in Transport Planning: Advances in Dynamic Traffic Assignment (Eds.: C. M. J. Tampere, F. Viti and L. H. Immers), Edward Elgar Publishing. pdf

4. Friesz, T. L., C. Kwon and D. Bernstein (2009), Evolutionary and Preferential Attachment Models of Demand Growth, in Complexity and Spatial Networks: In Search of Simplicity (Eds.: Aura Reggiani and Roberto Patuelli), Springer-Verlag. doi pdf

5. Friesz, T. L., C. Kwon and D. Bernstein (2007), Analytical Dynamic Traffic Assignment Models in Handbook of Transport Modelling 2nd Edition (Eds.: D. A. Hensher and K. J. Button), Elsevier. pdf

6. Friesz, T. L. and C. Kwon (2007), Strategic Freight Network Planning Models and a Dynamic Oligopolistic Urban Freight Network in Handbook of Transport Modelling 2nd Edition (Eds.: D. A. Hensher and K. J. Button), Elsevier. pdf

Edited Book

1. Batta, R. and C. Kwon (2013), Handbook of OR/MS Models in Hazardous Materials Transportation, International Series in Operations Research & Management Science, Vol. 193, Springer doi

Book

1. 엄태웅, 최윤섭, 권창현 (2019), 대학원생 때 알았더라면 좋았을 것들, 384쪽, ISBN-10: 1189430207, ISBN-13: 9791189430207.

2. C. Kwon (2016), Julia Programming for Operations Research: A Primer on Computing, 7.5" x 9.25". 246 pages. ISBN-10: 153332879X, ISBN-13: 978-1533328793. url