In online food delivery operations, accurate delivery time prediction is crucial, as it underpins effective resource allocation and ensures customer satisfaction. While prediction models can be trained using historical delivery data, there is significant room for improving accuracy, especially in incorporating driver-specific information. However, this task faces several challenges: (1) limited data availability and privacy concerns, (2) driver heterogeneity and data dispersion, and (3) tacit driver knowledge and feature engineering complexity. To tackle these issues, we introduce a Deep & Cross Network-based (DCN-based) framework. This framework utilizes limited driver-specific information and dispersed data to automate feature engineering and enhance prediction accuracy, enabling a more personalized and precise prediction process. It leverages both low-order feature interactions, captured by the Cross Network, and high-order interactions from the Deep Neural Network (DNN), effectively balancing interpretability and predictive power. Extensive experiments using real-world data from Zomato demonstrate that our approach with driver-specific information significantly outperforms both traditional and state-of-the-art models, achieving superior results across all regression accuracy metrics. The best performance yields a root mean square error (RMSE) of 3.6660, representing a 35.93% improvement over models without driver-specific information. Furthermore, the framework’s automatic feature engineering provides deeper insights into the interactions between driver information and external factors, offering a valuable tool for improving online food delivery operations.

Moment-based distributionally robust optimization (DRO) provides an optimization framework to integrate statistical information with traditional optimization approaches. Under this framework, one assumes that the underlying joint distribution of random parameters runs in a distributional ambiguity set constructed by moment information and makes decisions against the worst-case distribution within the set. Although most moment-based DRO problems can be reformulated as semidefinite programming (SDP) problems that can be solved in polynomial time, solving high-dimensional SDPs is still time-consuming. Unlike existing approximation approaches that first reduce the dimensionality of random parameters and then solve the approximated SDPs, we propose an optimized dimensionality reduction (ODR) approach. We first show that the ranks of the matrices in the SDP reformulations are small, by which we are then motivated to integrate the dimensionality reduction of random parameters with the subsequent optimization problems. Such integration enables two outer and one inner approximations of the original problem, all of which are low-dimensional SDPs that can be solved efficiently. More importantly, these approximations can theoretically achieve the optimal value of the original high-dimensional SDPs. As these approximations are nonconvex SDPs, we develop modified Alternating Direction Method of Multipliers (ADMM) algorithms to solve them efficiently. We demonstrate the effectiveness of our proposed ODR approach and algorithm in solving two practical problems. Numerical results show significant advantages of our approach on the computational time and solution quality over the three best possible benchmark approaches. Our approach can obtain an optimal or near-optimal (mostly within 0.1%) solution and reduce the computational time by up to three orders of magnitude.

Recently, autonomous driving system (ADS) has been widely adopted due to its potential to enhance travel convenience and alleviate traffic congestion, thereby improving the driving experience for consumers and creating lucrative opportunities for manufacturers. With the advancement of data sensing and control technologies, the reliability of ADS and the purchase intentions of consumers are continually evolving, presenting challenges for manufacturers in promotion and pricing decisions. To address this issue, we develop a two-stage game-theoretical model to characterize the decision-making processes of manufacturers and consumers before and after a technology upgrade. Considering the unique structural characteristics of ADS, which consists of driving software and its supporting hardware (SSH), we propose different business strategies for SSH (bundle or unbundle with the vehicle) and driving software (perpetual licensing or subscription) from the manufacturer's perspective. We find that, first, SSH strategies influence the optimal software strategies by changing the consumers' entry barriers to the ADS market. Specifically, for manufacturers with mature ADS technology, the bundle strategy provides consumers with a lower entry barrier by integrating SSH, making the flexible subscription model a dominant strategy; while perpetual licensing outperforms under the unbundle strategy. Second, the software strategies influence the optimal SSH strategy by altering consumers' exit barriers. Perpetual licensing imposes higher exit barriers; when combined with a bundle strategy that lowers entry barriers, it becomes a more advantageous choice for manufacturers with mature ADS technology. In contrast, the subscription strategy allows consumers to easily exit the market, making the bundle strategy advantageous only when a substantial proportion of consumers are compatible with ADS.

The full lifecycle concept has prompted sellers to provide ancillary services or products based on traditional product sales, leading to a dual-market system consisting of a base market and an add-on market. In this study, we consider a manufacturer selling base products through an online retail platform and then selling add-on products directly to consumers who have purchased base products. We investigate how the manufacturer’s distribution channel strategy in the base market interacts with the platform’s entry strategy in the add-on market. Results show that under the reselling (agency) channel, the platform’s entry of the add-on market enables the manufacturer to increase the wholesale price (reduce selling quantities to enjoy a higher margin) in the base market. We call this wholesale price effect (sales-control effect) caused by the platform’s entry. If the manufacturer adopts the agency (reselling) channel in the base market, the platform prefers (not) to enter the add-on market to compete with the manufacturer; if the manufacturer adopts the dual-channel, the platform enters only if both the commission rate and channel competition are high. Furthermore, the manufacturer prefers the dual-channel when both the commission rate and channel competition are low. Interestingly, due to the interactions between the two firms, the manufacturer will adopt the agency channel instead when the commission rate is extremely high. Finally, we examine conditions under which the platform has incentives to allow the manufacturer to change from a single-channel to a dual-channel in the base market.

This study proposes a mixed-integer programming formulation to model the individual-based path (IPR) recommendation problem during public transit service disruptions with the objective of minimizing system travel time and respecting passengers' path choice preferences. Passengers' behavior uncertainty in path choices given recommendations is also considered. We model the behavior uncertainty based on the passenger's prior preferences and posterior path choice probability distribution with two new concepts: epsilon-feasibility and Gamma-concentration, which control the mean and variance of path flows in the optimization problem. We show that these two concepts can be seen as a way of approximating the recourse function (expected system travel time) in a two-stage stochastic optimization. It is proved that these two concepts help to bound the difference between the approximated recourse function and the exact one. Additional theoretical analysis shows that epsilon-feasibility and Gamma-concentration can be seen as an approximation of expectation and chance constraints in a typical stochastic optimization formulation, respectively. The proposed IPR problem with behavior uncertainty is solved efficiently with Benders decomposition. The model is implemented in the Chicago Transit Authority (CTA) system with a real-world urban rail disruption as the case study. Results show that the proposed IPR model significantly reduces the average travel times compared to the status quo and outperforms the capacity-based benchmark path recommendation strategy.

We consider a network inventory system motivated by one-way, on-demand vehicle sharing services. Under uncertain and correlated network demand, the service operator periodically repositions vehicles to match a fixed supply with spatial customer demand while minimizing costs. Finding an optimal repositioning policy in such a general inventory network is analytically and computationally challenging. We introduce a base-stock repositioning policy as a multidimensional generalization of the classical inventory rule to n locations, and we establish its asymptotic optimality under two practically relevant regimes. We present exact reformulations that enable efficient computation of the best base-stock policy in an offline setting with historical data. In the online setting, we illustrate the challenges of learning with censored data in networked systems through a regret lower bound analysis and by demonstrating the suboptimality of alternative algorithmic approaches. We propose a Surrogate Optimization and Adaptive Repositioning algorithm and prove that it attains an optimal regret of O(n^2.5√(T)), which matches the regret lower bound in T with polynomial dependence on n. Our work highlights the critical role of inventory repositioning in the viability of shared mobility businesses and illuminates the inherent challenges posed by data and network complexity. Our results demonstrate that simple, interpretable policies, such as the state-independent base-stock policies we analyze, can provide significant practical value and achieve near-optimal performance.

As a promising application of cobots in labor-intensive warehouses, human-robot collaborative order picking systems provide a flexible and human-friendly picking solution by capitalizing on the best attributes of human pickers and robots. Few studies have determined operation modes of human-robot collaborative order picking systems to be beneficial to efficiency, cost, and the well-being of human workers. We identify four human-robot collaborative modes for order picking: single robot to single picker (couple), single robot to multiple pickers (SR-to-MP), single picker to multiple robots (SP-to-MR), and multiple pickers to multiple robots. For each mode, we establish a fork-join queuing network (FJQN) model to analyze system performance and apply a fatigue-recovery model to estimate the fatigue of the pickers. The proposed FJQN and fatigue-recovery model are validated by simulation. Although the throughput time and picker fatigue in the SR-to-MP mode can benefit from an appropriate zoning policy, we find, interestingly, that the zoning policy cannot reduce the throughput time in the SP-to-MR mode. The SP-to-MR mode is economical if a warehouse does not pursue a swift throughput time. A well-capitalized warehouse can adopt the SR-to-MP mode to improve the throughput time further in a more human-friendly manner.

The pretrain-transfer paradigm, which underpins the success of large language models (LLMs), has demonstrated the immense power of creating foundation models that learn generalizable representations from vast datasets. However, extending this paradigm to Operations Research (OR) problems on graph structures remains challenging due to the fundamental conflict between the statistical flexibility of language and the strict combinatorial constraints of graphs. To bridge this gap, we introduce the Graph Foundation Model (GFM), the first framework capable of solving all distance-based optimization problems on graph structures. By introducing the LLM-like self-supervised pre-training paradigm on the paths generated from random walks in the graph, GFM is compelled to internalize the graph's complex topological and combinatorial rules, where the connectivity of the structure itself can be treated as the supervisory signal. Unlike existing neural methods that learn complex and task-specific solving policies, our approach leverages the pre-trained GFM as a foundational model of the graph's intrinsic structure, which in turn enables a simple generative heuristic to tackle a diverse range of optimization challenges effectively. Comprehensive experiments on networks ranging from 20 to 893 nodes demonstrate that GFM achieves competitive performance against specialized solvers across a variety of distinct optimization task classes, while maintaining significantly faster inference times. Our work establishes a new paradigm of adapting the pretrain-transfer framework to graph optimization, opening the door for applying foundation model innovations to OR.

Contemporary society faces a growing set of complex issues representing significant socioeconomic, health and well-being, environmental, and sustainability challenges. The discipline of industrial and systems engineering (ISE) can play an important role in addressing these issues. This paper identifies and discusses eight grand challenges for ISE. These grand challenges are (1) Artificial Intelligence (AI) For Business and Personal Use: Decision-Making and System Design and Operations, (2) Cybersecurity and Resilience, (3) Sustainability: Environment, Energy and Infrastructure, (4) Health Issues, (5) Social Issues, (6) Logistics and Supply Chain, (7) System Integration and Operations: Humans, Automation, and AI, and (8) Industrial and Systems Engineering Education. The discussed grand challenges were derived by accomplished ISE professionals who are the authors of this paper. The implications of the ISE grand challenges for education, training, research, and implementation of ISE principles and methodologies for the benefit of global society are discussed.

Revenue management decisions often involve both offline and online decisions. Offline decisions are made first and establish the broad and long-term operational context in which online decisions are frequently and repeatedly made, often in real time. We consider a joint optimization of offline and online decisions. Specifically, we examine a setting in which the offline decision concerns the selection of product-design characteristics (e.g., price, capacity, return eligibility, and other characteristics) and the online decision concerns the dynamic assortment optimization over a selling season. Our formulation has many applications, including optimizing products’ return eligibility and determining product discounts, and a key feature of our model is its explicit consideration of complex return dynamics and accompanying financial implications. We formulate an optimization problem that combines the impact of both offline and online decisions on the expected revenue. To determine the product design, we reformulate the choice-based deterministic linear program, solve its continuous relaxation, and round the resulting solution. Using value function approximations enables us to obtain a dynamic assortment policy whose expected revenue is at least a constant fraction of the choice-based deterministic linear program. Combining these two results, we show that our approach provides an approximate solution to the joint optimization problem with performance guarantees. Numerical experiments based on real transaction data from a major U.S. retailer show that our method achieves 95%–97% effectiveness, an advantage of up to 18% over methods that disregard the interplay between offline and online decisions. This framework also yields a systematic quantitative measure of the relative importance of both offline and online decisions. Based on this measure, numerical experiments highlight the crucial role of product design, accounting for 94% and 85% of the observed variation in effectiveness across various methods in applications involving volume discount and return eligibility, respectively. This paper was accepted by Victor Martinez de Albeniz, operations management. Funding: This work was supported by the National Science Foundation’s Division of Civil, Mechanical and Manufacturing Innovation [Grant 2226901]. Supplemental Material: The online appendix and data files are available at https://doi.org/10.1287/mnsc.2022.01167 .

This paper presents data-driven approaches for integrated assortment planning and inventory allocation that significantly improve fulfillment efficiency at JD.com, a leading e-commerce company. JD.com uses a two-level distribution network that includes regional distribution centers (RDCs) and front distribution centers (FDCs). Selecting products to stock at FDCs and then, optimizing daily inventory allocation from RDCs to FDCs are critical to improving fulfillment efficiency, which is crucial for enhancing customer experiences. For assortment planning, we propose efficient algorithms to maximize the number of orders that can be fulfilled by FDCs (local fulfillment). For inventory allocation, we develop a novel end-to-end algorithm that integrates forecasting, optimization, and simulation to minimize lost sales and inventory-transfer costs. Numerical experiments demonstrate that our methods outperform existing approaches, increasing local order fulfillment rates by 0.54%, and our inventory allocation algorithm increases FDC demand satisfaction rates by 1.05%. Considering the high-volume operations of JD.com, with millions of weekly orders per region, these improvements yield substantial benefits beyond the company's established supply chain system. Implementation across JD.com's network has reduced costs, improved stock availability, and increased local order fulfillment rates for millions of orders annually.

Nonprogressive diffusion models the spread of behavior on social networks, where agents are allowed to reverse their decisions as time evolves. To provide an efficient framework for evaluating and optimizing nonprogressive diffusion, we introduce a comprehensive model along with a fixed-point approximation (FPA) scheme, which admits both theoretical guarantee and computational efficiency. We show that the approximation error depends on the network structure and derive order-optimal bounds for this error based on a newly proposed network measure. Additionally, we propose two easy-to-calculate network metrics (one at the node level and the other at the network level) that serve as reliable indicators of FPA performance. Our results indicate that the FPA scheme is particularly accurate for dense and large networks, which are typically challenging to analyze via simulation. To showcase the broad applicability of our approach, we apply the FPA scheme to well-known problems, like influence maximization and optimal pricing on social networks. Finally, we conduct extensive numerical experiments on both synthetic and real-world networks. On real-world networks, the FPA scheme achieves computational speedups of 70–230 times compared with naïve agent-based simulation and 23–30 times compared with a more advanced simulation method while maintaining a mean absolute percentage error of less than 3.48%. This paper was accepted by Jeannette Song, operations management. Funding: R. Zhang is grateful for financial support from the National Natural Science Foundation of China [Grant 72422004] and the Hong Kong Research Grants Council General Research Fund [Grants 14502722, 14504123, and 14503224]. Supplemental Material: The online appendix and data files are available at https://doi.org/10.1287/mnsc.2022.03031 .

Supply chain planning is the critical process of anticipating future demand and coordinating operational activities across the logistics network. However, within the context of contemporary e-commerce, traditional planning paradigms, typically characterized by fragmented processes and static optimization, prove inadequate in addressing dynamic demand, organizational silos, and the complexity of multi-stage coordination. To address these challenges, this study proposes a fundamental rethinking of supply chain planning, redefining it not merely as a computational task, but as an interactive, integrated, and automated cognitive process. This new paradigm emphasizes the organic unification of human strategic intent with adaptive execution, shifting the focus from rigid control to continuous, intelligent orchestration. To operationalize this conceptual shift, we introduce a Generative AI-powered agentic framework. Functioning as an intelligent cognitive interface, this framework bridges the gap between unstructured business contexts and structured analytical workflows, enabling the system to comprehend complex semantics and coordinate decisions across organizational boundaries. We demonstrate the empirical validity of this approach within JD.com's large-scale operations. The deployment confirms the efficacy of this cognitive paradigm, yielding an approximate 22

Time series neural networks perform exceptionally well in real-world applications but encounter challenges such as limited scalability, poor generalization, and suboptimal zero-shot performance. Inspired by large language models, there is interest in developing large time series models (LTM) to address these issues. However, current methods struggle with training complexity, adapting human feedback, and achieving high predictive accuracy. We introduce TimeHF, a novel pipeline for creating LTMs with 6 billion parameters, incorporating human feedback. We use patch convolutional embedding to capture long time series information and design a human feedback mechanism called time-series policy optimization. Deployed in JD.com's supply chain, TimeHF handles automated replenishment for over 20,000 products, improving prediction accuracy by 33.21 shows significant industrial benefits.

This study investigates the operations of a novel service model, mobile charging, in which an e-commerce platform operator dispatches trucks equipped with charging piles to provide charging services for customers' electric vehicles with low battery levels. We model the platform-customer interaction with a Stackelberg game and explicitly characterize customers' optimal charging decisions under the platform's various service plans. In a general scenario that involves charging requests within a transportation network, we develop a joint optimization model for the platform's pricing and service fulfillment, utilizing an elaborately constructed augmented network. In addition, we explore a localized subproblem where multiple orders are concentrated within a specific region. With a simplified model structure, we propose an approximation algorithm with provable performance guarantees and further theoretically evaluate the resource consumption and associated platform benefits for serving the orders in each region. The results enable the batching of neighboring demands within the general scenario as a specialized node, resulting in a streamlined network with fewer nodes. Additionally, we can enhance the algorithm's efficiency within the general framework through strategic prioritization of node visitations, leveraging the analytical findings. Furthermore, the insights derived can offer recommendations for the deployment of mobile charging and the selection of target areas in the initial stages. Overall, our study provides comprehensive guidelines and valuable insights for mobile charging operations.

We consider assortment and inventory planning problems with dynamic stockout-based substitution effects, and without replenishment, in two different settings: (1) Customers can see all available products when they arrive, a typical scenario in physical stores. (2) The seller can choose to offer a subset of available products to each customer, which is more common on online platforms. Both settings are known to be computationally challenging, and the current approximation algorithms for the two settings are quite different. We develop a unified algorithm framework under the MNL choice model for both settings. Our algorithms improve on the state-of-the-art algorithms in terms of approximation guarantee and runtime, and the ability to manage uncertainty in the total number of customers and handle more complex constraints. In the process, we establish various novel properties of dynamic assortment planning (for the MNL choice model) that may be useful more broadly.

In this work, we consider an online robust Markov Decision Process (MDP) where we have the information of finitely many prototypes of the underlying transition kernel. We consider an adaptively updated ambiguity set of the prototypes and propose an algorithm that efficiently identifies the true underlying transition kernel while guaranteeing the performance of the corresponding robust policy. To be more specific, we provide a sublinear regret of the subsequent optimal robust policy. We also provide an early stopping mechanism and a worst-case performance bound of the value function. In numerical experiments, we demonstrate that our method outperforms existing approaches, particularly in the early stage with limited data. This work contributes to robust MDPs by considering possible prior information about the underlying transition probability and online learning, offering both theoretical insights and practical algorithms for improved decision-making under uncertainty.

Drone swarms coupled with data intelligence can be the future of wildfire fighting. However, drone swarm firefighting faces enormous challenges, such as the highly complex environmental conditions in wildfire scenes, the highly dynamic nature of wildfire spread, and the significant computational complexity of drone swarm operations. We develop a predict-then-optimize approach to address these challenges to enable effective drone swarm firefighting. First, we construct wildfire spread prediction convex neural network (Convex-NN) models based on real wildfire data. Then, we propose a mixed-integer programming (MIP) model coupled with dynamic programming (DP) to enable efficient drone swarm task planning. We further use chance-constrained robust optimization (CCRO) to ensure robust firefighting performances under varying situations. The formulated model is solved efficiently using Benders Decomposition and Branch-and-Cut algorithms. After 75 simulated wildfire environments training, the MIP+CCRO approach shows the best performance among several testing sets, reducing movements by 37.3% compared to the plain MIP. It also significantly outperformed the GA baseline, which often failed to fully extinguish the fire. Eventually, we will conduct real-world fire spread and quenching experiments in the next stage for further validation.

Problem definition: We consider intertemporal pricing in the presence of reference effects and consumer heterogeneity. Our research question encompasses how to estimate heterogeneous consumer reference effects from data and how to efficiently compute the optimal pricing policy. Academic/practical relevance: Understanding reference effects is essential for designing pricing policies in modern retailing. Our work contributes to this area by incorporating consumer heterogeneity under arbitrary distributions. Methodology: We propose a mixed logit demand model that allows arbitrary joint distributions of valuations, responsiveness to prices, and responsiveness to reference prices among consumers. We use a nonparametric estimation method to learn consumer heterogeneity from transaction data. Further, we formulate the pricing optimization as an infinite horizon dynamic programming problem and solve it by applying a modified policy iteration algorithm. Results: Moreover, we investigate the structure of optimal pricing policies and prove the suboptimality of constant pricing policies even when all consumers are loss-averse according to the classical definition. Our numerical studies show that our estimation and optimization framework improves the expected revenue of retailers via accounting for heterogeneity. We validate our model using real data from JD.com, a large E-commerce retailer, and find empirical evidence of consumer heterogeneity. Managerial implications: In practice, ignoring consumer heterogeneity may lead to a significant loss of revenue. Furthermore, heterogeneous reference effect offers a strong motive for promotions and price fluctuations.

Download This Paper Open PDF in Browser Add Paper to My Library Share: Permalink Using these links will ensure access to this page indefinitely Copy URL Copy DOI