Abstract Edge AI refers to the deployment and execution of artificial intelligence (AI) algorithms on decentralized, local computing devices (e.g., Internet of Things) rather than relying on centralized cloud‐based computing infrastructures. The literature has addressed the increasing importance of AI in supply chain and operations management (SCOM), specifically in learning‐based, generative, and agentic AI, as well as digital twins. However, the area of Edge AI has not yet received a theoretical framing and conceptualization in SCOM. In this paper, we discuss the impacts of Edge AI on SCOM. We begin with an analysis of AI in SCOM, focusing on learning‐based, generative, agentic AI, and digital twins. We then detail Edge AI's impacts on SCOM across Digital SCOR functions: orchestration, fulfillment and ordering, planning, sourcing, regeneration, synchronization, and transformation. We also analyze organizational transformations that Edge AI might trigger. Specifically, we discuss interactions of agentic AI and Edge AI, an integration of local and centralized decision‐making, and deployment scenarios of Edge AI use in SCOM, focusing on production planning and control, intralogistics, and supply chain resilience. Risks and challenges of the new technology are considered as well. Impacts of Edge AI on supply chain and operations research are discussed. We conclude that Edge AI is a crucial part of the unfolding Industry 6.0 revolution that, in combination with agentic AI, provides a technology to implement decentralized, autonomously planned and controlled supply chain, production, and logistics processes, materializing in Industry 6.0 with AI across all layers of value‐added chains.
Abstract The rapid growth of civil aviation increases operational pressure on airport baggage handling systems. While robotic arms offer an automated alternative to manual processes, optimizing item placement for both loading efficiency and space utilization remains a critical challenge. This study formulates an online three‐dimensional loading problem to maximize trailer space utilization under operational constraints, including weight‐distribution constraints and pressure limits. We develop a batch‐based optimization algorithm with a physical buffer zone to dynamically group incoming baggage. The method integrates single‐bin serial and dual‐bin parallel loading modes with fixed and adaptive grouping strategies, supported by a staged multi‐objective evaluation function for sequence adjustment. Computational experiments show that the dual‐bin parallel mode with adaptive grouping achieves the best performance, improving average space utilization to 70.64% and outperforming the first‐in‐first‐out and OnlineBPH baselines. The proposed approach provides an effective and practical solution for automated loading operations in logistics and warehousing systems.
Abstract Food banks play a key role in reducing food insecurity, yet their operations are often constrained by manual and non‐optimised allocation procedures. This study introduces a methodological advance through an extended goal programming model with poverty incorporation (EGP‐PI). The EGP‐PI extends classical goal programming (GP) by embedding dynamically defined poverty measures directly into the deviation structure and by transforming multidimensional vulnerability attributes into endogenous optimisation weights. This creates a poverty‐sensitive, data‐driven prioritisation mechanism not present in traditional GP. The formulation also integrates dynamic inventory management and First Expired, First Out (FEFO)‐based shelf‐life control, ensuring coherence between perishability, fairness and efficiency across sequential connected planning periods. The model is validated using operational data from the São Paulo municipal food bank, one of the largest metropolitan systems in the world. Results show substantial improvements in allocation fairness, poverty targeting and resource utilisation, providing a quantitative and socially responsive tool for humanitarian food‐distribution systems.
Abstract In modern manufacturing systems, integrating preventive maintenance into production scheduling is essential to ensure operational efficiency and equipment reliability. This paper addresses the flexible job shop scheduling problem under machine unavailability constraints caused by non‐fixed maintenance tasks. We propose four enhanced mixed‐integer programming models subject to maintenance tasks in flexible job shops. We present four formulation enhancements that significantly improve the performance of the mathematical models. In addition, an efficient constraint programming model is introduced. We introduce a comprehensive dataset of 2010 instances derived and extended from classical flexible job‐shop scheduling benchmarks. The proposed benchmark covers a wide range of configurations, making it suitable for evaluating solution robustness and scalability. An extensive computational study demonstrates the strengths and trade‐offs among the proposed models. The computational experiments show the superiority of the constraint programming formulation compared to mixed‐integer programming models. This work proposes both efficient modeling approaches and a valuable benchmark for future research.
Abstract This work aims at minimizing the energy purchase costs related to the operation of a green hydrogen production plant connected both to an intermittent, uncertain renewable energy source and to the power grid. We focus on scenarios where, for economic or environmental reasons, the plant must primarily meet hydrogen demand using electricity from wind power, despite its uncertain availability. The production problem is represented as a lot‐sizing model, and an original hybrid robust‐stochastic approach is developed to consider the time‐correlated uncertainty despite a limited knowledge of its probability distribution. The developed approach is computationally efficient and further accelerated thanks to an in‐depth study of the generated robust sub‐problem. Simulations on a realistic industrial problem demonstrate the benefits of the presented approach.
