ICORES 2026 Abstracts

Area 1 - Methodologies and Technologies
Full Papers Short Papers
Area 2 - Applications
Full Papers Short Papers

Area 1 - Methodologies and Technologies

Full Papers
Paper Nr: 25
Title:

A Bidding-Collaborative Model for Weapon-Target Assignment against Integrated Defense Systems
Authors:

Xiaozhan Li, Yuanhang Li, Yufan Deng, Jianing Li and Guangquan Cheng
Abstract: To address the challenges in the Weapon-Target Assignment (WTA) problem—such as heterogeneous combat unit capabilities, large decision spaces, and the difficulty of real-time decision-making—this paper proposes a Bidding-Collaborative Fire Strike (BCFS) mode inspired by market mechanisms. The BCFS mode introduces a “task publication→bid submission→order dispatch” workflow, where the bidding process is formulated as a binary programming model. This model translates different command strategies into corresponding objective functions, balancing decision accuracy and timeliness. Under a realistic combat scenario that incorporates coordinated strikes by Red Force brigades against Blue Force radar stations, air defense/missile interception units, and ground-based jamming equipment, we constructed test cases with varying target distributions and problem scales. Using the Gurobi solver, optimal strike plans under two distinct strategies are generated. Experimental results demonstrate that the BCFS mode can efficiently solve small-scale WTA problems under dynamic target coordination scenarios. Moreover, changes in command strategy are clearly reflected in the resulting strike plans, indicating the model’s adaptability to multiple tactical strategies.
Download
Paper Nr: 27
Title:

Evaluation of ICU Bed Reservation Policies Using Loss Models with IPP Arrivals: Balancing Efficiency and Fairness
Authors:

Wei Tian, Anqi Wang, Hanzhi Zhang and Jingjin Wu
Abstract: The effective allocation of Intensive Care Unit (ICU) beds is critical for balancing urgent emergency admissions and elective cases, particularly during demand surges. This paper develops a loss queuing model with a threshold-based bed reservation policy to evaluate the quality of service in terms of patient rejection rates. Unlike prior studies that assume Poisson arrivals, we apply the Interrupted Poisson Process (IPP) to capture the bursty nature of realistic patient flows. Using real ICU data, we demonstrate that IPP provides a more accurate fit to arrival patterns compared with Poisson processes. Analytical derivations and discrete-event simulations are employed to evaluate class-specific and overall rejection rates under different reservation strategies. Results show that reserving more beds for emergency patients reduces their rejection rates but increases those for elective patients. Consequently, this compromises service fairness across patient groups. We adopt two approaches to ensure fairness: 1) optimal control to determine the reservation threshold, and 2) an alternative priority mechanism for allocating beds to emergency patients when both reserved and common beds are available. We further demonstrate that rejection rates remain nearly insensitive to LoS distribution beyond its mean, which broadens the applicability to diverse real-world scenarios. The findings provide practical guidance for hospitals in determining dynamic and fair reservation thresholds, emphasizing the necessity of dynamic bed allocation policies.
Download
Paper Nr: 29
Title:

Energy-Efficient Scheduling in Parallel Machines with Speed Scaling and Release Dates
Authors:

Ahmed Missaoui and Barry O'Sullivan
Abstract: The manufacturing sector ranks as the second-largest energy consumer, following electricity generation. With fluctuating energy prices and significant reliance on fossil fuels, operative energy management has become a pressing challenge for the manufacturing sector to remain competitive and reduce operational costs. To address these issues, manufacturers are increasingly adopting advanced energy-efficient technologies and optimizing energy-efficient scheduling practices. This study focuses on the parallel machine scheduling problem with release date, aiming to minimize both the makespan and total energy consumption. A mixed-integer linear programming (MILP) model is formulated, and the augmented epsilon-constraints method is used to derive the optimal Pareto front for small-scale problem instances. Furthermore, two efficient multi-objective approaches based on iterated greedy and iterated local search are developed to solve a benchmark set of 200 instances. In a comprehensive computational study, both methods demonstrate strong performance.
Download
Paper Nr: 40
Title:

Bi-Objective Electric Vehicle Charging Scheduling with Stochastic Vehicle Arrivals
Authors:

Aimen Khiar, Mohamed el Amine Brahmia and Lhassane Idoumghar
Abstract: Electric vehicles present a sustainable alternative to conventional transportation due to their reduced environmental impact and enhanced energy efficiency. However, their prolonged charging times create significant operational hurdles, motivating the development of intelligent scheduling systems that optimize charging resources and improve operational efficiency. In this paper, we address the constrained multi-objective stochastic scheduling problem for EV charging operations, where vehicle arrivals are uncertain; specifically, clients may cancel their requests with a given probability. The objective is to minimize both the expected mean relative tardiness and the expected peak load, thereby addressing charging station stability and client satisfaction simultaneously. This approach offers a more realistic framework compared to existing state-of-the-art methods for this problem. We demonstrate that computing the expected peak load for a given schedule requires an exponential number of operations with respect to the problem size. Given this computational complexity, we adopt a Monte Carlo approximation approach and employ metaheuristic algorithms to approximate the optimal Pareto front using problem-specific operators designed to explore the solution space effectively. Specifically, we implement both versions of the Non-Dominated Sorting Genetic Algorithm (NSGA-II and NSGA-III) along with the Multi-Objective Cuckoo Search (MOCS) algorithm. We conduct numerical experiments comparing these metaheuristics in terms of hypervolume and execution time, demonstrating that the MOCS algorithm yields the best performance. Furthermore, we demonstrate that our stochastic model substantially outperforms its deterministic counterpart by explicitly accounting for demand uncertainty, resulting in more robust and operationally efficient EV charging schedules that better reflect real-world variability and establish the practical superiority of this work.
Download
Paper Nr: 45
Title:

Row and Column Generation Approaches for a Three-Stage Industrial Packing Problem for Truck-Transport Preparation
Authors:

Khadija Hadj Salem, Benoit Lardeux, Xavier Schepler and Tony Wauters
Abstract: Large retailers need advanced supply chain solutions to ensure their products are always available to customers. One of the challenges in this context is packing products for transport by truck from distribution centers to stores. This process involves placing box types (homogeneous or heterogeneous) on pallets, stacking them, and then loading and balancing the stacks into trucks. These steps ensure that products are transported at the lowest possible cost to the transportation company. Motivated by a major global retailer's need to solve this problem, we introduce the three-stage packing problem (3-stage PP) with practical packing constraints. The problem is first formulated as a mixed-integer linear program, and the computation of the derived upper and lower bounds is given. Two decomposition approaches are then proposed. The first approach relies on row (constraint) generation, while the second relies on column generation. The results of computational experiments with randomly generated, realistic instances are reported. These results allow us to evaluate the effectiveness of the proposed approaches.
Download
Paper Nr: 46
Title:

Inventory Policies for a Lindley-Type Stochastic System under Partial Observation
Authors:

Hugo Cruz-Suárez and Ruy López-Ríos
Abstract: This article studies a Lindley-type inventory system in which demand is only partially observed. The manager observes realized sales but receives no information on unmet demand whenever stockouts occur. The objective is to minimize the expected discounted total cost through suitable ordering (or production) strategies. To address this partially observable setting, we formulate the problem using unnormalized demand densities, which yield a linear belief recursion and a tractable dynamic programming equation. The Bellman operator is analyzed and shown to be valid under this representation, leading to the existence of optimal policies. Furthermore, we establish the connection between value functions defined with unnormalized beliefs and those based on normalized beliefs, clarifying the equivalence of both formulations. A numerical example illustrates the approach and demonstrates the appearance of (s,S)-type policies.
Download
Paper Nr: 61
Title:

A Constraint Programming Model for the Super-Agile Earth Observation Satellite Imaging Scheduling Problem
Authors:

Margarida Caleiras, Samuel Moniz and Paulo Jorge Nascimento
Abstract: As the dependence on satellite imaging continues to grow, modern satellites have become increasingly agile, with the new generation, namely super-agile Earth observation satellites (SAEOS), providing unprecedented imaging flexibility. The highly dynamic capabilities of these satellites introduce additional challenges to the scheduling of observation tasks, as existing approaches for conventional agile satellites do not account for variable observation durations and multiple imaging directions. Although some efforts have been made in this regard, the SAEOS imaging scheduling problem (SAEOS-ISP) remains largely unexplored, and no exact approaches have yet been proposed. In this context, this study presents the first exact Constraint Programming formulation for the SAEOS-ISP, considering flexible observation windows, multiple pointing directions and sequence-dependent transition times across multiple satellites. Computational experiments on a newly generated benchmark set demonstrate that the model can be solved efficiently and within very short computational times. Moreover, the results also show that the proposed approach has the potential to achieve higher computational performance compared to the non-exact approaches that are currently considered state-of-the-art.
Download
Paper Nr: 63
Title:

Joint Optimization of Production and Condition-Based Maintenance with Speed-Dependent Degradation
Authors:

Maëlys Durrieu, Jean-Philippe Gayon, Alex Kosgodagan-Dalla Torre and Guillaume Massonnet
Abstract: Maintenance operations and production parameters are often optimized separately. We consider a system in which the production speed directly influences the system’s degradation. We propose a combined approach to jointly optimize the production speed and the maintenance decisions under a condition-based maintenance policy in order to maximize the average production rate. We show that in a deterministic setting, the optimal strategy consists in operating at a constant production speed until the system reaches the selected degradation threshold. For the specific case where the degradation rate depends on a power of the production speed, we derive a closed-form expression for the optimal speed as well as a simple optimization procedure for determining the optimal degradation threshold. In that case we study the effect of several parameters on the optimal policy and production rate.
Download
Paper Nr: 65
Title:

A Mixed-Integer Linear Programming Model for Job-Shop Scheduling with Flexible Cooperative Robot Transportation
Authors:

Jean-Philippe Gayon, Philippe Lacomme and Amine Oussama
Abstract: This paper introduces a new extension of the Job-Shop Scheduling Problem with Transportation Resources (JSPT), called the Job-Shop Scheduling Problem with Flexible Cooperative Robot Transportation (JSPT-FC), which integrates cooperative robots capable of dynamically forming teams to execute transportation tasks. Each transportation task is performed by a team of robots whose size and composition are decision variables. The model relies on a speedup assumption, in which larger cooperative teams achieve shorter transportation times through shared load handling, but at the cost of synchronization delays and reduced overall system parallelism. A Mixed-Integer Linear Programming (MILP) formulation is proposed to jointly determine, for each transportation task, the robot team’s size and composition, as well as the production and transportation schedules, under robot synchronization and cooperation constraints. The formulation is evaluated on adapted JSPT benchmark instances extended to the JSPT-FC framework. Computational experiments with two speedup scenarios demonstrate that robot team size flexibility significantly influences makespan and solver behaviour, greatly increasing problem complexity. The results reveal a strong trade-off between transportation speedup achieved through robot cooperation, the loss of system parallelism, and the synchronization delays induced by larger robot teams.
Download
Paper Nr: 71
Title:

Reinforcement Learning Methods for Neighborhood Selection in Local Search
Authors:

Yannick Molinghen, Augustin Delecluse, Renaud De Landtsheer and Stefano Michelini
Abstract: Reinforcement learning has recently gained traction as a means to improve combinatorial optimization methods, yet its effectiveness within local search metaheuristics specifically remains comparatively underexamined. In this study, we evaluate a range of reinforcement learning-based neighborhood selection strategies – multiarmed bandits (upper confidence bound, ε-greedy) and deep reinforcement learning methods (proximal policy optimization, double deep Q-network) – and compare them against multiple baselines across three different problems: the traveling salesman problem, the pickup and delivery problem with time windows, and the car sequencing problem. We show how search-specific characteristics, particularly large variations in cost due to constraint violation penalties, necessitate carefully designed reward functions to provide stable and informative learning signals. Our extensive experiments reveal that algorithm performance varies substantially across problems, although that ε-greedy consistently ranks among the best performers. In contrast, the computational overhead of deep reinforcement learning approaches only makes them competitive with a substantially longer runtime. These findings highlight both the promise and the practical limitations of deep reinforcement learning in local search.
Download
Paper Nr: 89
Title:

Two-Moment Phase-Type Fitting for ASIP Tandem Systems
Authors:

Wesley Geelen and Maria Vlasiou
Abstract: The asymmetric simple inclusion process (ASIP) is an n-site tandem stochastic network used to model clustered particle flow. Particles at each site form a cluster that moves to the next site when its gate opens. For the general ASIP, no closed-form expression exists for the load in terms of the first two moments of the interarrival and inter-gate opening time distributions. We approximate the steady-state load by fitting a phase-type ASIP using a two-moment approach. We obtain a recursive expression for the site-occupancy PGF and a closed-form PGF for the steady-state load, enabling a closed-form approximation of its expectation. Our results show that interarrival and inter-gate opening distributions have opposite effects on the load. Further, contrary to typical performance evaluation of queuing systems, our results indicate that systems with more unreliable servers (increased inter-gate opening time variance) exhibit superior expected performance, assuming that the servers’ expected performance is equivalent.
Download
Paper Nr: 106
Title:

GS-Mitigate: An Implemented Group-Stratified Decision-Support Framework for Pandemic Mitigation Outcomes
Authors:

Anita Tadakamalla and Alexander Brodsky
Abstract: Decision-makers require analytical tools to design pandemic mitigation strategies that balance health outcomes with operational and productivity costs. While classical compartmental epidemiological models are analytically tractable, they often lack the structural resolution needed to support policy trade-off analysis across interacting interventions. This paper presents GT-Mitigate, a fully implemented, group-stratified epidemiological modeling framework that extends prior SUEIHCDR-based models for decision-oriented analysis of comprehensive mitigation protocols. The population is represented as homogeneous groups defined by multiple attributes: extended epidemiological state, vaccination status, and intervention adoption. Disease progression and mitigation effects are modeled through deterministic, rule-based group-wise transitions, eliminating probability estimation and enabling transparent computation of health, cost, and productivity key performance indicators. Illustrative simulation experiments demonstrate how alternative policy configurations influence epidemic trajectories and operational burdens, revealing fundamental trade-offs and providing a foundation for future optimization-based policy design.
Download
Short Papers
Paper Nr: 23
Title:

Radiotherapy Scheduling Under Patient Arrival Uncertainty
Authors:

Hugues Rauwel, Christian Artigues, Romain Guillaume and Laure Vieillevigne
Abstract: Radiotherapy scheduling is a challenging problem faced by many cancer treatment centers. It involves assigning and scheduling patient sessions based on available resources. Sub-optimal schedules can lead to service congestion and unnecessary patient waiting times. Maximizing the use of available knowledge during scheduling is crucial. In this paper, we propose a new two-stage model for the radiotherapy scheduling problem, accounting for uncertain patient arrivals (i.e. patients awaiting medical validation). We explore different strategies for incorporating patients with uncertain arrival times. Our evaluation is based on a simulation using real-world data over multiple scheduling days. We compare our two-stage approach against a deterministic method, where only the patients with fixed arrival time are scheduled. Results show that our scenario-based approach outperforms the deterministic one over time, leading to reduced patient waiting times. Additionally, we find that selecting a small number of representative scenarios is an effective strategy, given the computational complexity of solving large scenario sets. This approach demonstrates the potential for more efficient radiotherapy scheduling and ultimately improves patient care.
Download
Paper Nr: 32
Title:

Sequential Games with Turn Selection Process, Incomplete Information and Fuzzy Utility Functions
Authors:

Rubén Becerril-Borja
Abstract: A group of models of sequential games is studied where the order of the turns is not known beforehand by the players, and where the utility functions for each player are fuzzy numbers. Moreover, players can be of different types so their utility function changes and therefore players have incomplete information about the opponents. For these models, a series of results are proven to show the existence of equilibria, and a brief application is described where it is adequate to use fuzzy numbers due to the uncertainty of utilities.
Download
Paper Nr: 33
Title:

Mean–Variance Portfolio Optimization with Shrinkage Estimation for Recommender Systems
Authors:

Tomoya Yanagi, Yuta Yasumoto and Yuichi Takano
Abstract: Recommender systems are aimed at resolving the information overload problem by providing a personalized list of items that are unknown but attractive to each user. We consider a mean–variance portfolio optimization model with a cardinality constraint for generating high-quality lists of recommendations. The quality of recommendation lists can be improved by applying portfolio optimization. However, it is usually difficult to accurately estimate the rating covariance matrix required for mean–variance portfolio optimization because of a shortage of observed user ratings. To improve the accuracy of covariance matrix estimation, we apply shrinkage estimation methods that compute the weighted sum of the sample and target covariance matrices. We propose two types of target matrices that work well for shrinkage estimation of the rating covariance matrix. Experimental results show that with appropriate parameter tuning, our method can improve the quality of recommendation lists produced by collaborative filtering algorithms.
Download
Paper Nr: 36
Title:

Smaller Linear Programs from Algorithms via Constraint Sharing
Authors:

Shermin Khosravi and David Bremner
Abstract: Linear Programming (LP) has broad applications in industry and is foundational to many other optimization methods. Its expressiveness as a modeling tool has been the focus of extensive research. Recent advances introduced an LP compiler translating polynomial-time algorithms into polynomial-size LPs, even for polynomial-time problems having exponential extension complexity. This approach enables modeling in a high-level programming language, offering an alternative to Algebraic Modeling Languages (AMLs), which require users to explicitly define each set of constraints. Despite being polynomial in size, the LPs are large in practice, making them challenging for current solvers. As part of our broader goal of systematically producing Compact Integer Programs (CIPs) for exponential-size Integer Programs (IPs) having polynomial-time separation oracles, we propose a many-to-one mapping of identical modeling language statements into shared sets of constraints. We further enhance this approach with placeholder substitutions that increase similarity among constraint-heavy operations. Our reduction methods exploit program structure and LP design insights inaccessible to conventional presolve and compiler optimizations. We demonstrate the effectiveness of our approach on the makespan and matching problems, both having exponential extension complexity. The resulting LPs are up to 53% smaller and substantially easier to solve across commercial and open-source solvers.
Download
Paper Nr: 38
Title:

Exact and Metaheuristic Approaches for Unrelated Parallel Machine Scheduling with Sequence-Dependent Setups and Shared Resources
Authors:

Samuel Nucamendi-Guillén, Héctor G.-de-Alba, Oliver Avalos-Rosales and Francisco Ángel-Bello
Abstract: This paper addresses the unrelated parallel machine scheduling problem with sequence-dependent setup times and shared resources, aiming to minimize the makespan (Cmax). A mixed-integer linear programming (MILP) formulation and a Multi-Start (MS) algorithm combined with a Basic Variable Neighborhood Descent search strategy (BVND) are proposed to efficiently tackle small and medium-sized instances. The MILP model incorporates a valid inequality to strengthen the lower bound, while the MS+BVND approach integrates diversification mechanisms and parameter tuning procedures within its constructive phase to enhance exploration. Computational experiments show that the formulation optimally solved instances with up to 14 jobs and 4 machines within a one-hour computational limit. However, their efficiency decreases rapidly as the instance size increases. In contrast, the MS+BVND algorithm consistently provides near-optimal solutions with an average deviation of less than 5.18% inCmax compared to the MILP best-known solutions, while requiring less than 1% of the CPU time. Moreover, MS+BVND exhibits consistent and scalable behavior across medium instances, maintaining high solution quality and moderate computational effort. These results confirm that MS+BVND is a robust, efficient, and scalable metaheuristic alternative for complex scheduling problems where exact MILP model become computationally intractable.
Download
Paper Nr: 41
Title:

Constraint Programming and Simulated Annealing Approaches for Parallel-Machine Scheduling with Conflict Constraints, Server Setups, and Flexible Maintenance
Authors:

Rachid Benmansour, Angelo Sifaleras and Raca Todosijevic
Abstract: This paper addresses a parallel-machine scheduling problem where jobs require setups performed by a single server and must respect conflict constraints that prevent certain jobs from running simultaneously. This type of problem can find applications in logistics and transport operations, particularly when scheduling vehicle fleets that share limited resources. The server is also subject to a fixed-duration maintenance activity that must be scheduled alongside the jobs. The objective is to minimize the makespan. We develop both constraint programming and simulated annealing approaches to solve this problem. Experimental results demonstrate that the constraint programming model, executed on the Minizinc solver, successfully obtains optimal solutions for small instances with 10 jobs. For larger problems with 15 jobs, simulated annealing is a good alternative since it allows to obtain solutions at 2.3% of the optimum on average despite a fixed calculation time of 10 seconds.
Download
Paper Nr: 42
Title:

Scheduling a Single-Robot Manufacturing Cell Under Blocking Constraints: A MILP-Based Decomposition Approach
Authors:

Eduardo Redondo and Clara Ruiz Diaz
Abstract: We study a non-preemptive blocking job-shop scheduling problem motivated by a single-robot welding cell. We develop a mixed-integer linear programming (MILP) formulation that models blocking, robot-holding modes, and travel times within a unified structure. On small and moderate instances, the monolithic MILP attains exact solutions; beyond roughly 10–15 jobs, runtimes increase due to the interaction between machine sequencing and robot synchronization. Computational difficulty does not follow a monotonic relation with workload or makespan, indicating the combinatorial effect of blocking constraints. To extend scalability and include urgent arrivals without modifying ongoing operations, we introduce a MILP-based block decomposition that partitions the job list into sequential subproblems of calibrated size. Blocks are solved in a sort–partition–MILP pipeline-first ordered by due dates (EDD) and then by processing times (SPT)-and concatenated while preserving feasibility. Experiments show that the decomposition produces feasible schedules while reducing total computation time and enabling predictable reoptimization latency. The framework can be applied to offline planning in single-robot manufacturing systems and to reactive rescheduling when urgent jobs occur.
Download
Paper Nr: 48
Title:

A Dual Framework for Corporate Carbon Management: Aligning the Locus of Emission and the Locus of Control
Authors:

Wassim Amzil and Ahlam Azzamouri
Abstract: Decarbonization has become a key goal for modern organizations due to regulatory pressure, investor expectations, and the urgency of climate change. However, even with significant investments and clear ambitions, many large firms still struggle to achieve meaningful emission reductions. This challenge arises more from the complexity of corporate structures that obscure accountability for decisions related to emissions than from a lack of resources. In many cases, it remains unclear who truly has the authority to influence or stop emissions. This disconnect leads to what we call the attribution fallacy, a gap between where emissions occur and who actually controls them. To address this, the study introduces the Dual-attribution network framework, which links emission sources to the departments that influence them. It combines Life Cycle Assessment (LCA) showing where emissions occur in physical processes, and Input-Output (IO) analysis explaining how decisions across departments or sectors affect them. The framework operates in two steps: mapping physical sources of emissions, then identifying who controls them using a Distributed Control Index (DCI). By turning emission maps into accountability maps, the framework turns decarbonization from isolated sustainability projects into a shared managerial responsibility and a daily organizational practice embedded across all functions.
Download
Paper Nr: 49
Title:

Beyond the Vehicle Routing Problem: Design of Temporal Networks for Demand-Responsive Transport
Authors:

Xiaoyi Wu, Ravi Seshadri, Filipe Rodrigues, Carlos Lima Azevedo and Andrea Araldo
Abstract: Conventional public transportation (CPT) is composed of fixed routes and fixed timetables, usually determined via long-term planning, based on nominal demand. However, during operations, demand may greatly deviate from the nominal one, causing a mismatch between demand and supply, leading to an inefficient service. On the other hand, flexible mobility services, such as Demand-Responsive Transport (DRT), adapt bus routes to the actual user demand. However, routes are calculated by solving a Vehicle Routing Problems (VRPs), which are not as effective as CPT in terms of demand consolidation, resulting in cost inefficiency. While in CPT, consolidation is obtained by forcing users to adapt to CPT by lines, VRP adapts instead to bus routes to user demand. This work introduces an alternative approach to DRT operations: different from VRP, we design a structured network describing bus routes, allowing for complex user trips, including transfers and walking legs. This enables greater consolidation and efficiency. While network design problems are limited to static networks, we propose here an original formulation to design temporal networks, which allows structured bus routes to adapt to the observed demand. We provide a proof-of-concept of the proposed approach, and show in small-scale numerical experiments that it reduces operator cost, without excessively penalizing users, compared to the classic VRP-based solution.
Download
Paper Nr: 53
Title:

A Genetic Algorithm Approach to the Generalized Assignment Problem with Non-Linear Costs for Humanitarian Supply Chains
Authors:

Paraskevas Dimitriou, Vasileios Karyotis and Panos Kourouthanassis
Abstract: Humanitarian supply chains (HSC) have emerged as critical factors for crisis situations, e.g., floodings, earthquakes, etc. Although similar in concept to traditional supply chains at delivering goods to specific locations, HSCs operate under high uncertainty and different constraints/requirements. In this paper, we study a HSC system as a generalized assignment problem with non-linear costs in the objective function, in an attempt to address the problem more realistically. Due to inherent computational complexity by the involved costs, we propose a new Genetic Algorithm for solving the problem more efficiently. We detail its design, analyze its complexity and compare it against a mixed-integer linear programming with piecewise-linear approximation approach and a refined piecewise-linear formulation that explicitly enforces a Special Ordered Set of type 2 (SOS2) structure using binary activation constraints. Our results demonstrate the efficacy of our approach as the size of the problem instance increases and show great promise in real-world problems.
Download
Paper Nr: 55
Title:

A Framework to Develop a Demonstrator for Cyber-Physical System
Authors:

Faizan Ahmed, Pieter Zeilstra, Sebastian Piest and Jeroen Linssen
Abstract: This paper proposes a methodological framework designed to provide guidelines for building a demonstrator for various tasks within the context of cyber-physical systems. The framework is based on a design science framework that provides a systematic methodology to help reshape the objectives, attributes, and functioning of the system. The proposed framework is evaluated using a real-world case study demonstrating the practical value of applying the framework in industrial settings. By correctly identifying stakeholders, setting clear objectives, and capturing the necessary business knowledge, the resulting solution became not only technically feasible but also meaningful and usable for its intended users.
Download
Paper Nr: 57
Title:

Half-Open Multi-Depot Electric Vehicle Routing Problem with Time-Dependent Stochastic Waiting Times at Recharging Station
Authors:

Hakan Erdeş and Saadettin Erhan Kesen
Abstract: This paper addresses Half-Open Multi-Depot Electric Vehicle Routing Problem with Time-Dependent Stochastic Waiting Times at Recharging Station (HO-MDEVRP-TDSWT-RS) where departure and arrival depots for EVs may be different. While EVs can obtain recharging services from depots without waiting, they may wait in queue at the RSs for recharging. We consider that RSs are equipped with multiple chargers and follow a stochastic nature based on 𝑀/𝑀/𝑐 queueing system depending on the periods of the day. We solve this problem in two stages, each formulated as Mixed Integer Linear Programming (MILP) model. In the first stage, expected waiting times at RSs for any period is calculated and the problem is solved accordingly. In the succeeding stage, waiting time for EVs upon arrival at the RS is generated and the problem is solved considering realized waiting time and customers following the RS. In this case, customers can be served by both the related EV and newly dispatched EVs. We conduct experiments with instances derived from recent literature. Results indicate that new EVs are dispatched in more than one third of all instances in the second stage and our bespoke recourse action can reduce the total cost by more than 30% in certain instances.
Download
Paper Nr: 64
Title:

From Readiness to Action: An Operational Research Framework Supporting Portuguese SME Digital Transformation
Authors:

Rui Pinto and Gil Gonçalves
Abstract: Small and Medium-Sized Enterprises (SMEs) play a critical role in economic growth, yet struggle to adopt digital transformation due to limited resources, lack of expertise, and operational constraints. Existing maturity models are either too generic or fail to provide actionable guidance tailored to SMEs. This paper proposes a structured self-assessment framework that evaluates digital maturity across seven dimensions and generates practical technological and strategic recommendations. The framework was developed using a design-science research approach combining a systematic literature review with empirical data collected from SMEs through a quantitative survey and iterative user testing. The resulting tool enables SMEs to assess readiness, prioritize initiatives, and align digitalization strategy with operational goals. Validation results show that SMEs score higher in digital culture and openness to transformation but remain significantly underdeveloped in data analytics, cybersecurity, and Industry 4.0 technologies. Feedback from pilot usage confirms the framework’s usability and relevance, with 87% of respondents stating it supports decision-making. The proposed framework bridges the gap between theory and practice and contributes a practical, replicable artifact for guiding SME digital transformation.
Download
Paper Nr: 66
Title:

Numerical Analysis of Reverse Quantum Annealing in Domain-Wall Encoded Continuous-Variable Optimization
Authors:

Jiei Iwama and Hironori Makino
Abstract: Domain-Wall Encoding (DWE) is one of the implementation techniques that enables the handling of continuous variables in quantum annealing (QA)-based quantum computers, and is expected to expand the applicability of QA to a broader range of optimization problems. This paper investigates, from a numerical perspective, the effectiveness of Reverse Quantum Annealing (RQA)-a refinement technique for solution extraction via QA-when applied to the process of deriving optimal solutions represented by continuous variables. The results reveal the existence of a characteristic time scale that maximizes the effect of RQA. These findings not only pave the way for theoretical analysis of non-adiabatic transitions between energy eigenstates through quantum interference effects, but also provide important insights for designing effective operational schedules in DWE-based QA machines.
Download
Paper Nr: 69
Title:

Finite Sample Maximum Delays in a Queue with Deterministic Service
Authors:

Pierre M. Fiorini
Abstract: We study the maximum waiting/response time attained during a busy period in the M/D/1 queue when the system operates near capacity (ρ ≈ 1), with a focus on finite-horizon, finite-sample extremes rather than asymptotic limits. Let MW denote the supremum of the workload process W(t) over a busy period of random length T. Using renewal arguments, we derive a Volterra-type renewal equation for the conditional CDF FM(x | w) = P(MW (w) ≤ x) and solve it numerically via discretization and fixed-point iteration. This busy-period CDF is then used in order-statistic formulas to obtain exact finite-sample expectations E[M[n] W ] and tail probabilities over n i.i.d. busy periods (or, equivalently, over a finite observation window with N arrivals), where M[n]W denotes the maximum of the n busy-period maxima. Numerical results for ρ = 0.99 show near-logarithmic growth E[M[n]W ] ≈ a(ρ) + b(ρ)logn on practical ranges of n, with b(ρ) consistent with the Cramer–Lundberg exponent. We finally illustrate how these finite-sample distributions can be used directly in ´capacity planning under strict worst-case SLA constraints.
Download
Paper Nr: 80
Title:

Reinforcement Learning Algorithms in Operative Management of Railway Nodes
Authors:

Andrea Galadíková, Norbert Adamko and Jozef Kostolný
Abstract: Providing decision support in a railway node can be challenging, especially at the operational level, as the railway node is a complex system with many unpredictable influences, such as dynamic scheduling changes, varying train arrival times, and unexpected maintenance issues affecting its operation. When addressing operational decision support in a railway node, it is necessary to explore the new possibilities of using artificial intelligence and machine learning. In this paper, we examine ways to facilitate the decision-making process of dispatchers using reinforcement learning algorithms. A methodology was proposed, detailing the implementation process of the algorithm to ensure its applicability for decision support. This methodology ensures that the solution is practical and scalable. We decided to apply this approach to a specific task: determining maintenance activities in a maintenance depot, an important yet complex component of the railway network. A reinforcement learning algorithm, namely proximal policy optimization, was used, and the results suggest that this approach may be a viable option for supporting decision-making process.
Download
Paper Nr: 81
Title:

Improved MILP Formulation for Home Healthcare Scheduling and Routing with Multiple Depots
Authors:

Ifeoma Ogbodo, Wasakorn Laesanklang and Dario Landa-Silva
Abstract: Despite extensive research on home healthcare scheduling and routing problems (HHCSRP), a critical gap persists between mathematically optimal solutions and operationally feasible implementations. This paper demonstrates that constraints often considered redundant in vehicle routing formulations are essential for HHCSRP, where worker-depot assignments cannot be arbitrarily changed. We validate a widely cited multi-depot HHCSRP MILP formulation using 42 real-world instances, revealing that 77.8% of optimal solutions contain operational violations, workers incorrectly assigned to arbitrary depots and unproductive direct depot-to-depot routes without patient visits. Our main contribution is a refined formulation with explicit operational feasibility constraints that eliminate these violations, while improving computational efficiency on average by 40%. Comparative analysis using GUROBI and CPLEX solvers reveals instance-dependent performance patterns, with GUROBI achieving faster solving times for small to medium resource-constrained instances, and CPLEX producing superior solutions for large-scale, over-resourced problems. These findings underscore that operational validation must extend beyond standard optimisation metrics to verify real-world practicability, a persistent gap contributing to the scarcity of successful HHCSRP deployments in practice.
Download
Paper Nr: 82
Title:

Extended MILP Formulation to Solve Home Healthcare Scheduling and Routing Problem with Unpaid Caregiving
Authors:

Ifeoma Ogbodo and Dario Landa-Silva
Abstract: Traditional home healthcare scheduling and routing problems (HHCSRP) have primarily focused on paid caregivers, overlooking the significant contribution of unpaid caregivers (family members and friends) who provide approximately GBP 162 billion worth of care annually in England and Wales. This paper presents an extended mixed-integer linear programming (MILP) formulation that explicitly integrates unpaid care provision as decision variables within HHCSRP, incorporating medically informed eligibility constraints for service tasks. Computational experiments using 42 modified real-world instances demonstrate significant operational gains, with average cost savings of 59.45% at the highest level of unpaid care provision. Through systematic analysis across different provision scenarios (0%, 10%, 30%, 50%, 70%, 90%), we identify an optimal provision level of 30%, yielding the highest marginal benefit (20.92% additional improvement over the 10% provision level), while balancing solution quality and computational tractability. Interestingly, integrating unpaid care constraints enhances computational efficiency, with several complex instances reaching near optimality in seconds instead of hours. The findings demonstrate that explicitly modelling unpaid care constraints can improve the mathematical structure of HHCSRP as well as provide cost benefits.
Download
Paper Nr: 93
Title:

Queueing Model with Alternating Service for Zipper Merging
Authors:

Yuki Goto and Tuan Phung-Duc
Abstract: Efficient traffic flow management is critical for ensuring safety, optimizing logistics, and promoting environmental conservation. Merging points are areas where congestion and accidents are likely to occur, requiring a reconsideration of merging strategies. While “zipper merging—a method where vehicles from the merging lane merge into the main lane, alternating one-by-one with main lane vehicles—”is known to be efficient, few studies have examined the capacity constraints of merging lanes. In this study, we propose a queueing model with two alternating service queues to analyze this method. We specifically focus on the impact of the finite buffer capacity of the merging lane (K) and the arrival rate of vehicles on the merging lane (λ2). Under the assumption of zero switchover times—defined as the time required for the server to switch service between queues—we derive the stability condition and the stationary distribution using a Quasi-Birth-and-Death (QBD) process. Our analysis reveals a critical managerial insight: while increasing the capacity K reduces the blocking probability for the merging lane, it effectively reduces the stability region of the main lane by increasing the server occupancy for the merging traffic. These findings provide a theoretical foundation for determining the optimal buffer capacity to balance traffic flow efficiency and system stability.
Download
Paper Nr: 95
Title:

Turning Re-Loyalty Scores into Action: A Mixed-Integer Optimization Framework for Broadband Offers in B2B Telecom
Authors:

Rafaela Maciel Caetano Alonso and Gladston Luiz da Silva
Abstract: Credit and service retention in business telecommunications remains challenging, particularly in mature B2B/SME markets where acquiring new contracts is costly. Although many operators deploy predictive models to estimate renewal probabilities, these scores are often not connected to a systematic decision process that selects offers under commercial constraints. This paper presents a prescriptive optimization framework that turns re-loyalty scores into actionable broadband recommendations. Using an existing Gradient Boosting model (LightGBM) that outputs a customer-level renewal probability, we formulate a mixed-integer program that selects an ordered Top–3 set of offers per eligible customer to maximize expected retention value. The model combines the event probability with the current billed amount and enforces capacity, eligibility, and pricing constraints, while excluding rarely observed price/velocity transitions via empirical behavioral rules. We evaluate the approach on real B2B/SME data from a large Brazilian telecom operator under two commercial policies: moderate price increases up to +50% and aggressive increases up to +100%. Results show that the optimization engine produces consistent Top–3 recommendations at scale and highlights the coverage–uplift trade-off across pricing strategies.
Download
Paper Nr: 96
Title:

Predictive Maintenance in South African Coal Mining: A Data-Driven Approach to Operational Efficiency
Authors:

Ritshidze Juarez Nethenzheni and Josef Langerman
Abstract: Coal mining in South Africa operates in harsh and dynamic environments, which makes equipment maintenance complex. Heavy machines such as continuous miners are exposed to extreme operating conditions, which increase the risk of breakdowns. Despite increasing interest in data-driven maintenance, industry practice remains largely reactive and preventive. This disconnect creates a clear gap in the knowledge and practical application of machine learning for predictive maintenance. Predictive maintenance (PdM) offers a strategic response to these challenges by using machine learning and sensor data to predict equipment failures. This research develops a context-specific PdM model for underground coal mining on a Joy Continuous Miner, using operational and telemetry data. The study applies the CRISP-DM framework to guide model development and testing. The research shows how ensemble models such as Extra Trees and XGBoost can effectively predict failure events, reducing unplanned downtime and improving maintenance scheduling accuracy. The findings show that producing interpretable predictive models which are context-specific can improve operational efficiency and asset reliability within underground mining operations. However, there are limitations, including data imbalance and noise from sensor variability.
Download
Paper Nr: 102
Title:

An Integrated Hybrid Framework for Green Heterogeneous Vehicle Routing with Mixed Pickup and Delivery
Authors:

Prateek Gupta, Devanand, Jorge Augusto Meira, Antonio Ken Iannillo, Daniel Antunes Pedrozo and Danilo D'Aversa
Abstract: In this paper, we address the Heterogeneous Vehicle Routing Problem with Mixed Pickup and Delivery (HVRPMPD), integrating fuel-based emission parameters for CO2e emission estimation. Our methodology focuses on developing a time-budgeted matheuristic framework capable of immediate deployment under strict real-world operational constraints, including heterogeneous fleets and work hours. Initial solutions are obtained using a greedy insertion heuristic with swap-based local improvements, modeling fuel consumption as a function of load, speed, and route factors. These solutions are then refined through a Mixed-Integer Linear Programming (MILP) formulation under a time limit to balance operational and environmental objectives. This structure supports planners in trading off runtime and emission reductions by enabling fast, swappable heuristic components that provide quick starting points and support refinement of the final solution. To quantify the environmental impact of shifting from standard commercial practices to specialized emission routing, we test on real-world data with 20 to 55 service points reflecting the typical daily operational scale of European logistics Small and Medium-sized Enterprises (SMEs). The results demonstrate that our framework presents a 15% reduction in emissions compared to the Google Routing API, which is frequently utilized by organizations for immediate, low-cost, first-version routing deployment and optimized for time or distance.
Download
Paper Nr: 103
Title:

Realistic Multi-Agent Path Planning in Constrained Industrial Environments Applied to New Task Execution Unit Benchmarks
Authors:

António R. Teixeira and Ana Moura
Abstract: This article introduces TEU (Task Execution Unit), a micro-factory–inspired benchmark for evaluating Multi-Agent Path Finding (MAPF) and Multi-Agent Pickup and Delivery (MAPD) algorithms under realistic industrial constraints. Motivated by the rapid expansion of MAPF research and large-scale AMR deployments, TEU addresses limitations of classical formulations that overlook continuous task streams, resource contention, congestion, and embodied robot motion. The benchmark models compact containerised environments as narrow grids with fixed material, workstation, and expedition cells, time-to-live (TTL) resource holds, and task dependencies executed by suspended autonomous mobile robots. Two layouts (3×5 and 4×10), are combined with 100 stochastic scenarios per layout, resulting in 200 fully reproducible benchmark instances. Three coordination approaches are evaluated: an idealised geometric lower bound (LB), a sequential priority-based (SPB), and a joint multi-robot space–time A* (JMR-ST-A*). Performance is assessed using makespan, travel distance, computation time, and the number of avoided collisions. Results demonstrate that spatial confinement and resource interactions dominate the system behaviour, revealing pronounced trade-offs between throughput and computational cost. TEU therefore provides a compact yet expressive benchmark to support future research in layout optimisation and learning-based multi-robot coordination.
Download
Paper Nr: 107
Title:

Integrating Production Uncertainty into Dairy Supply Chain Planning: A Stochastic Programming Model for Tunisia Case
Authors:

Inès Saidi, Rim Kalai and Mohamed Amine Abbessi
Abstract: Milk production in Tunisia exhibits marked seasonal variability, affecting the economic performance of collection systems. This issue primarily affects small dairy farmers, who have much to gain from finding new collaborative approaches to improve their incomes. A previous study proposed a collaboration model based on sharing refrigerated tanks to increase small farmers’ profits, improve quality, and limit milk waste. This article extends the initial model to incorporate uncertainty about the quantity of milk produced. A finite−scenario approach is adopted, defining four production states (high, medium, low, and very low) based on empirical data from recent literature on milk yield variability. The stochastic formulation maximizes expected profit under capacity and allocation constraints, with advance decisions on location and collection. The results show that accounting for uncertainty substantially alters the optimal network structure and yields more robust solutions than those of the deterministic model. This contribution provides an analytical framework for planning milk collection in environments subject to production variability.
Download
Paper Nr: 22
Title:

Accelerating Stochastic Optimization with Adaptive Lagrangian Cuts: Case Studies in Supply Chain and Network Flow
Authors:

Xiaoyu Luo, Mingming Xu and Chuanhou Gao
Abstract: Two-stage stochastic mixed-integer programs (SMIPs) are widely used to support decision-making under uncertainty. Classical Benders-type algorithms can be significantly strengthened with Lagrangian cuts, which provide tighter bounds but are notoriously expensive to generate. To address this challenge, we propose Adaptive Lagrangian Cuts (ALC), an improved cut generation scheme that preserves the strength of Lagrangian cuts while substantially reducing their computational burden. ALC accelerates cut generation by a factor of 2–5 and significantly decreases the branch-and-bound tree size. To demonstrate its practical relevance, we evaluate ALC on the stochastic server location problem (SSLP), representing supply chain facility location under uncertain demand, and the stochastic multicommodity flow problem (SMCF), modeling transportation and network flow planning with stochastic demands. Computational experiments confirm that ALC achieves scalable optimization performance across both domains, highlighting its potential as a practical enhancement of Benders-type algorithms for real-world stochastic optimization.
Download
Paper Nr: 47
Title:

A Structured Approach for Managing Security and Privacy Requirements Risk in Enterprise Systems
Authors:

Paulo Ribeiro, Tiago Pedrosa, Rui Ribeiro and Ricardo J. Machado
Abstract: Managing security and privacy risks in the design of complex enterprise systems is a critical challenge. Decision-makers often struggle to translate high-level strategic goals and regulatory constraints into verifiable operational controls, leading to systemic vulnerabilities and compliance gaps. This paper introduces and evaluates SPIDESOFT, a novel, two-phased risk management framework for designing complex enterprise systems. SPIDESOFT provides a structured, top-down process for transforming business goals and risk drivers into actionable specifications. The framework’s versatility is demonstrated through its application to four distinct functionalities within a large-scale mobility maintenance platform, each with a unique risk profile: real-time tracking (RTT), incident reporting (IR), predictive maintenance (PM), and parts inventory management (PIM). Our findings show that the method’s sequenced approach-which mandates early-phase data minimization and structured threat modeling-creates an explicit and traceable path from strategic intent to operational control design. The key contribution is a practical framework that makes security and privacy a traceable, evidence-based component of the enterprise system design process, enabling better risk management and resource allocation.
Download
Paper Nr: 51
Title:

Supply Chain Redesign Framework for Dual-Use of Industrial Materials for Energy Storage
Authors:

Youssef Makhlouf and Ahlam Azzamouri
Abstract: The mining sector, a global energy consumer, faces increasing pressure to adopt Renewable Energy Sources (RES). However, the inherent variability of RES necessitates Energy Storage Systems (ESS), which often impose significant capital expenditure (CAPEX), challenging their economic viability. This paper introduces a framework to leverage existing in-house materials through a redesigned supply chain to serve a dual function; as industrial resources and energy storage medium. The study outlines the required redesign of supply chain’s infrastructure and flows, integrating material adaptation and charge/discharge processes. A mathematical model is developed to analyse capacity degradation dynamics, with results validated through simulation. Finally, a stock and flow diagram establishes a foundational structure for future quantitative analysis. The proposed framework offers economic advantages by reducing upfront ESS investment through the utilization of materials and infrastructure already on-site, and by the continuous cycling of material which represents a mechanism to eliminate long-term capacity fading, even when using lower-performance materials. However, the feasibility of such framework is contingent on material-specific properties, technology costs, and potential impact on material quality. Overall, this study is intended to serve as a reference framework for future quantitative analyses and applications of the dual-use of industrial materials in energy storage systems.
Download
Paper Nr: 56
Title:

Sustainable Closed-Loop Supply Chains: A Systematic Literature Review
Authors:

Juan Orejel, Elías Olivares-Benitez, Abraham Mendoza and Gerardo Trevino-Garza
Abstract: This article presents a systematic literature review of sustainable Closed-Loop Supply Chains (CLSCs), emphasizing their transformative role in converting linear systems into environmentally and socially responsible models. The study synthesizes recent advancements in CLSC primarily from 2024 and 2025, highlighting their capacity to recover and reuse resources, reduce waste generation, and address social and environmental challenges. Building on foundational concepts introduced in closed-loop models, this paper highlights the evolving integration of sustainability principles with forward and reverse logistics processes. The review identifies significant research gaps, such as the need for models capable of managing uncertainties arising from natural disasters, technological disruptions, or global crises like the COVID-19 pandemic. Future research opportunities include enhancing optimization techniques to address these uncertainties, exploring the potential of advanced technologies in CLSC network design, and investigating emerging challenges such as fluctuating consumer behavior and carbon emissions regulations. By examining these critical areas, this study aims to advance the understanding of CLSC dynamics and support innovative solutions for a sustainable supply chain future.
Download
Paper Nr: 60
Title:

Better Queues, Better Care: The Q.U.E.U.E. Framework for Clinic Flow Decisions
Authors:

Gerald Jandusay
Abstract: Outpatient clinics have a unique conundrum of balancing two essential yet often conflicting priorities: workflow performance and patient experience. To improve clinic workflow and patient journey, this position paper introduces the QUEUE Framework, which is an acronym for five key factors: Queue Dynamics, User Interaction, Efficiency & Flow, Utilization of Resources, and Emotional Response. This practical approach integrates queueing theory with the psychology of queueing to support a multi-criteria evaluation of operational performance and experiential outcomes, grounded in healthcare operations management literature. It not only leads to more reliable and sustainable reductions in observed and perceived wait times, it also challenges the traditional approach in Operations Research (OR) which often simplifies assumptions for tractability. Thus, the QUEUE Framework improves upon conventional OR by introducing “Insight-First Modelling” to prioritize actionable insights, “Fairness-Constraint” to incorporate perceived fairness as a constraint and objective within the OR system, and lastly, “Interpretability by Design” to ensure generated insights are interpretable for non-OR members. This integrated model highlights the relevance and impact of applying the interaction between operational systems and the human element, extending beyond the healthcare industry to other service-oriented industries.
Download
Paper Nr: 62
Title:

Smart Supplier Selection in Facility Management through Multi-Criteria Analytics
Authors:

Nicolas Zante, Christophe Cruz and Sebti Foufou
Abstract: This paper addresses the supplier pre-selection problem in multi-site Facility Management (FM) procurement, where decision-makers must compose coherent supplier panels capable of jointly satisfying heterogeneous site requirements. It relies on a previously introduced mathematical optimization framework that systematically assigns suppliers to sites by modeling their capabilities and tender requirements within a unified multi-criteria characteristic space. A similarity-based scoring function is used to quantify the alignment between supplier offerings and site needs across multiple dimensions, including services, certifications, standards, and geographical constraints. Decision-maker preferences are incorporated through a dynamic weighting mechanism, enabling adaptive prioritization of criteria while preserving mathematical consistency. The resulting problem is formulated as a combinatorial optimization model inspired by the classical facility location problem, where similarity scores replace transportation costs. To efficiently solve large-scale instances, a simulated annealing metaheuristic is proposed. The algorithm explicitly defines solution representation, neighborhood structure, and acceptance criteria, allowing reproducible exploration of the solution space while balancing solution quality and computational effort. The algorithmic approach is evaluated through computational experiments on synthetic Facility Management datasets derived from a real-world supplier registry, involving up to 49,638 supplier agencies and problem instances reaching combinatorial sizes of up to 1017 possible assignments. Experimental results show that the simulated annealing algorithm consistently achieves near-optimal solutions with runtime much lower than exhaustive enumeration. These results demonstrate the scalability and practical relevance of the proposed algorithmic resolution as a decision-support tool for structured, transparent, and evidence-based supplier selection in complex multi-site FM environments.
Download
Paper Nr: 74
Title:

Optimizing Courier Positioning and Demand Coverage in Online Food Delivery Platforms
Authors:

Mohammadamin Tavasoli, Roberto Baldacci and Sara Ghanbari
Abstract: Online Food Delivery (OFD) platforms face the challenge of dynamically positioning couriers at waiting points to serve spatiotemporally varying demand at restaurants. This study develops a mixed integer programming model for the integrated problem of waiting point location, restaurant allocation, courier assignment, and relocation over a discrete planning horizon. The model incorporates practical operational constraints, including coverage radius limits, allowable relocation distance, courier capacity bounds, and waiting point capacity restrictions. To evaluate the impact of service flexibility, the model is formulated under two coverage paradigms: full coverage, which requires the complete fulfillment of assigned restaurants, and partial coverage, allowing for flexible demand fulfillment under capacity constraints. Computational experiments demonstrate that the partial coverage variant consistently achieves superior performance, with substantially higher service levels, using the same resources. Coverage radius emerges as the dominant factor in the solution, while relocation distance exhibits secondary effects.
Download
Paper Nr: 77
Title:

Optimizing Job Rotation in Assembly Lines by Balancing Productivity and Worker Well-Being
Authors:

Joana Rafaela Almeida, João Rafael Almeida and José Luís Oliveira
Abstract: Industry 4.0 technologies have improved productivity but also increased physical and cognitive demands on workers, particularly in manual assembly tasks. These increasing demands directly contribute to health problems, absenteeism, and reduced worker motivation. Traditional strategies to mitigate ergonomic risks, such as workstation redesign or task automation, are often costly or infeasible. Job rotation appears as a flexible alternative that distributes physical load across workers, being promotive of teamwork and workforce adaptability. This paper proposes a bi-objective mixed-integer linear programming (MILP) model for job rotation scheduling in assembly lines, integrating operator skills and medical restrictions to assign tasks considering different rotations in a shift. The model simultaneously minimizes (i) the maximum ergonomic exposure, quantified through metabolic energy expenditure, and (ii) the maximum cycle time, reflecting production efficiency. Ergonomic assessment accounts for task-specific characteristics and operator-specific factors, ensuring individualized workload considerations. A case study is conducted using data from a Portuguese boiler manufacturing line, evaluating the impact of varying operator counts and task assignments on ergonomic workload and line performance under different demand scenarios. Results show that the framework supports sustainable workforce management by enabling efficient task allocation, minimizing ergonomic risk and maximizing production efficiency. The code to validate this study is publicly available at https://github.com/j-rdalm/JobRotation.
Download
Paper Nr: 86
Title:

Pragmatic Approaches for Identifying and Exploiting Throughput Bottlenecks
Authors:

Pedro Costa, Ana Moura, Galina Robertsone and Tatjana Tambovceva
Abstract: Production efficiency is essential for gaining a competitive edge, and identifying bottlenecks is crucial for improving production systems. This study focuses on identifying and eliminating production constraints on a fire detector assembly line in a Portuguese company. To achieve this, the Process Time Method (PTM) was employed along with other tools to pinpoint inefficiencies. Using the PTM, the assembly line's constraints were identified by determining the resource with the longest effective process time. After pinpointing the bottleneck, specific interventions were implemented, including redefining the layout and standardizing workstations, to reduce the identified inefficiencies. The results demonstrated significant improvements, as the line capacity increased and the relative standard deviation (RSD) decreased notably, indicating enhanced process throughput and stability. This research advances the literature on manufacturing process improvement by formalizing a structured methodology that integrates the PTM with qualitative root-cause identification, data-driven prioritization and operational standardization. It demonstrates that systematic bottleneck management can achieve significant gains in process stability and throughput in data-scarce environments, providing a replicable methodology for industries with limited digital maturity.
Download
Paper Nr: 105
Title:

Hybrid GPU-Accelerated Pattern Generation for High-Performance CSP/BPP Optimization
Authors:

Francesca Guerriero and Francesco Paolo Saccomanno
Abstract: This paper proposes a novel GPU-Accelerated Pattern Generation approach that exploits the massive parallel architecture of GPUs to accelerate the pattern generation phase, bypassing the sequential limitations of traditional pricing. Unlike traditional Branch-and-Price methods, that seek the exact optimal pattern at every iteration, the proposed approach utilizes deterministic tree expansion to rapidly generate a diverse and high-quality pool of cutting patterns in parallel, and a restricted Set Covering formulation solved on the CPU to select the optimal combination of these patterns. We analyze the solution quality (optimality gap) and computational runtime on a set of diverse benchmark instances, ranging from standard uniform distributions to the challenging Falkenauer triplets. The hybrid heuristic successfully identifies the optimal bin solution for all instances with an average runtime of 0.6 seconds, significantly outperforming traditional approaches.
Download

Area 2 - Applications

Full Papers
Paper Nr: 24
Title:

Distributed Algorithms in the Field: Solving the Fusion Network Topology Selection Problem
Authors:

Imadeddine Aziez and Raphaël Boudreault
Abstract: The increasing complexity of modern defense and surveillance operations requires the integration of heterogeneous sensors and computing resources to deliver timely and accurate situational awareness. This paper introduces the Fusion Network Topology Selection Problem (FNTSP), a combinatorial, multi-objective, and constrained optimization problem that jointly addresses the allocation of fusion algorithms and the routing of data within a distributed network. We propose a mathematical formulation that captures the transformation of data through hierarchical fusion processes, while incorporating additional valid constraints and variable fixing techniques to reduce the search space. We evaluate the model’s performance on a benchmark of varied instances inspired by a realistic Canadian Northern use case. Our experiments demonstrate that the proposed approach can optimally solve small-size instances and generate feasible solutions for larger and more complex networks. These findings provide insights into the design of scalable and resilient distributed fusion systems, and establish a foundation for future research on advanced, more practical solution methods.
Download
Paper Nr: 28
Title:

CCH Ecosystem: Automated Carrier Communication System for European Digital Logistics
Authors:

Jorge Augusto Meira, Evgeny Polyachenko, Carlos Andre Zavadinack, Gabriel Irribarem Soares Ruas, Nikolai Riumin, Daniel Antunes Pedrozo, Renato Silva de Melo, Radu State and Kristof Horvat
Abstract: In this paper, we present Carrier Communication Hub (CCH), a production-ready platform designed to automate carrier relationship management for logistics providers. The system supports over 750 carriers across multiple markets and serves over 2,700 active customers, solving the inefficiencies of manual communication in large-scale operations. CCH operates in two modes: direct order placement and competitive bidding auctions. Carrier selection is automated through a combination of geographic filters, configurable groups, and intelligent scoring mechanisms. Communication is based on structured email templates that capture web-based responses, with instant messaging extensions for coordination at critical moments. The platform leverages machine learning techniques to enable dynamic pricing, allowing real-time adaptation to customer-specific requirements and integration with the CRM system. CCH uses historical transaction data to process delivery opportunities at scale. Furthermore, we propose a recommendation engine that evaluates carriers using a weighted scoring based on three configurable factors: recent activity, price competitiveness, and route experience, with additional adjustments for fleet capacity, geographic relevance, and annual inflation to ensure fair comparisons. By integrating automated communication workflows, an intelligent carrier recommendation engine, and dynamic pricing models, CCH enables scalable capacity expansion without a proportional increase in the workforce, offering a robust data-driven solution for managing complex logistics operations.
Download
Paper Nr: 44
Title:

Evaluating Operational Efficiency Trends in India's Major Ports: A Data Envelopment Analysis (DEA) with Variable Returns to Scale
Authors:

Pankaj Kumar Detwal and Rajat Agrawal
Abstract: This research assesses the operational efficiency of India’s major ports over the period 2005–2019 using an input-oriented, Banker, Charnes, and Cooper (BCC) Data Envelopment Analysis (DEA) model with Variable Returns to Scale (VRS). While earlier studies assess port efficiency at isolated points in time, systematic longitudinal evidence remains limited. This work examines trends over time and measures each port's performance against an efficient frontier derived from empirical data. The analysis uses a dataset compiled from the annual reports of the Ministry of Ports, Shipping and Waterways, available on the Open Government Data (OGD) platform. Findings indicate that while a few ports consistently operate on the frontier, most reach best practice only intermittently. A comparison of average efficiency scores before and after 2014 underscores the significant impact of port modernization initiatives. Moreover, the frequent emergence of specific ports as benchmarks reveals their lasting competitive advantages and broader influence across the system. For ports identified as inefficient, the slack analysis provided precise figures for potential input reductions without affecting output. The extended timeframe and a DEA approach customized to account for sectoral variations bring fresh insights, offering practical recommendations for port managers on enhancing performance and optimizing resource allocation.
Download
Paper Nr: 76
Title:

The Electric Vehicle Routing Problem with Hard Time Windows and Nonlinear Charging and Discharging
Authors:

Kai Yoshida, Ken Kobayashi, Kazuma Kawai, Yutaro Ito, Noriaki Ikemoto and Kazuhide Nakata
Abstract: This study addresses the electric vehicle routing problem (EVRP) with hard time windows, nonlinear charging, and load-dependent discharging. Most existing works rely on simplified assumptions, such as linearized charging, failing to adequately capture the critical interaction between battery characteristics and customer time windows. This limits their applicability to real-world operations. To overcome this limitation, we first formulate a path-based model that avoids duplicating charging stations while flexibly incorporating piecewise-linear approximations. This formulation enables the integrated optimization of routing and charging schedules under hard time windows. To tackle large-scale instances, we propose a heuristic framework based on Adaptive Large Neighborhood Search (ALNS) enhanced with Simulated Annealing. The method incorporates a tailored initial solution generation method and novel neighborhood operators. These operators are specifically designed to exploit the problem’s characteristics, addressing both nonlinear charging inefficiencies and the total route duration objective by manipulating time-window structures. Computational experiments on benchmark instances demonstrate that the proposed method yields high-quality solutions within short computation times. An ablation study further confirms the significant contribution of the new operators to the heuristic’s performance.
Download
Paper Nr: 99
Title:

A Three-Stage Lexicographic Constraint Programming Approach for an Energy-Efficient Scheduling Problem
Authors:

Pietro Girardis, Mirko Cavecchia and Manuel Iori
Abstract: The increasing heterogeneity of energy sources and the rising demand for energy are pushing industrial systems toward more sophisticated scheduling strategies. This paper addresses an energy-efficient scheduling problem motivated by a real-world consortium involved in the storage, distribution, and sale of apples. The problem concerns the coordination of daily refrigeration cycles for multiple cooling cells, whose availability is subject to stochastic variation. Each cell must alternate between cooling and idle phases while complying with technological and operational constraints. The system is modeled as an Energy-Efficient Identical Parallel Machine Scheduling Problem with precedence constraints and three objectives: minimizing peak energy consumption; achieving a balanced temporal allocation of cooling cycles; and ensuring robustness with respect to all possible availability scenarios. A three-stage lexicographic Constraint Programming approach is proposed, with each stage optimizing one objective while preserving the solutions obtained in the previous stage. Computational experiments based on real operational data show that the proposed approach provides solutions that have good quality on all considered objectives, by requiring a limited computational effort.
Download
Short Papers
Paper Nr: 18
Title:

Analysis and Optimization of Multi-Battery Management in Telecommunications Networks Within Retail and Curtailment Electricity Markets
Authors:

Julien Khamphousone, Mustapha Bouhtou, Matthieu Chardy, Youssouf Hadhbi and Camille Richer
Abstract: This paper addresses the optimization of multi-battery systems across multiple sites within telecommunications networks, specifically under the constraints of retail (called also SPOT) and curtailment electricity markets. We present two optimization approaches to solve this problem, which is defined as a combinatorial optimization problem that we prove to be NP-Hard. We also emphasize the importance of battery usage in generating savings and maximizing profits. First, we propose an Integer Linear Programming (ILP) formulation to optimize battery utilization while minimizing operational costs and maximizing overall profitability. Additionally, a greedy heuristic is introduced as an alternative solution approach, designed to provide quicker, near-optimal solutions suitable for real-sized instances. Extensive computational experiments using real data from the French telecommunications operator Orange are conducted to assess the performance of our proposed approaches.
Download
Paper Nr: 19
Title:

Evolving Teaching Methods in Operations Management: A Longitudinal Study for Industry 4.0 in Higher Education
Authors:

Rafael Henríquez Machado, Andrés Felipe Muñoz Villamizar and Lina Rocío Ramos Pachón
Abstract: This paper presents a longitudinal study that explores the impact of evolving teaching methods on student learning outcomes and teaching evaluations in an undergraduate Operations Management course. Over nine years, data were collected from 2,847 students at a Latin American university. The study traces the instructional shift from traditional lecture-based teaching to gamified learning activities and, more recently, to the integration of social media as a pedagogical tool. Findings show that student learning outcomes improved from basic to very competent levels, while teaching evaluations increased from 4.72 to 4.93 out of 5.0. These results suggest that gamification and social media, when used strategically, enhance student engagement, conceptual understanding, and perceptions of teaching effectiveness. The study contributes to the growing body of research advocating for innovative, student-centered approaches in production and operations education to meet the demands of Industry 4.0.
Download
Paper Nr: 26
Title:

Enhancing Noisy Inventory Data Accuracy in Perishable Supply Chains Using a Fast and Robust Observer Algorithm
Authors:

Valentina Orsini
Abstract: Reliable inventory data is an essential prerequisite for an effective management of supply chains. The purpose of this paper is to propose an efficient, cost-effective alternative to the widely used conventional methods based on high-cost sensor technologies. The proposed new approach relies on integrating a physical sensor with a custom-built estimation algorithm designed to derive accurate inventory data from inherently noisy sensor readings. The two key advantages are: an accurate, low-cost inventory estimation despite measurement noise, the capacity to regulate the convergence speed of the inventory estimation error toward zero.
Download
Paper Nr: 35
Title:

Fault-Tolerant Task Scheduling in Grid Topology
Authors:

Vaibhav Prabhu, Abhishek Mishra and Kamal Sheel Mishra
Abstract: Fault-tolerant task scheduling is critical in distributed systems where processor failures may significantly degrade performance. This paper studies fault-tolerant scheduling of directed acyclic task graphs (DAGs) on two-dimensional grid topologies. We present a contention-aware event-queue simulation model that incorporates task duplication using integer processor duplication mappings, referred to as ( f,g)-mappings, to tolerate single-processor failures. Five duplication strategies-diagonal, random, inverse, sinusoidal, and logarithmic-are evaluated on randomly generated DAGs of sizes up to 400 nodes. Performance is measured using Normalized Schedule Length (NSL), defined as the ratio of makespan to the critical path length. Experimental results consistently show that logarithmic mapping achieves lower NSL by balancing communication locality and contention reduction. We discuss modelling assumptions, limitations, and directions for future work.
Download
Paper Nr: 39
Title:

Probabilistic Risk Assessment Framework for Nascent Supply Chain Demonstrated through the Case of CBG in India Integrating Synthetic and Real Data
Authors:

Anindita Sarker and Rajat Agrawal
Abstract: The resilience of Bio-Compressed Natural Gas (Bio-CNG)/Compressed Biogas (CBG) Supply Chain is pivotal for India's target to achieve net-zero by 2070. Under the "Galvanizing Organic Bio-Agro Resources Dhan (GOBARdhan)" initiative, the Indian Government is pushing to establish a Bio-CNG industry. Despite recent growth, the sector remains in a nascent phase, exposed to multifaceted risks. Bio-CNG in India is produced from diverse feedstocks such as Napier grass, agricultural residue, poultry droppings, and municipal organic waste. The sustainability of this emerging sector is subject to various interconnected risks, including feedstock availability, quality, production, stubble burning, geographical constraints, and transportation disruptions, underscoring the requirement for risk assessment. The paper proposes a Bayesian-Monte Carlo framework for probabilistic risk assessment. The methodology involves defining a Bayesian Network structure, incorporating expert knowledge through structured elicitation to quantify uncertainties, utilising Monte Carlo simulation for dynamic risk analysis and validation, and the construction of a comprehensive risk matrix to aid decision-makers. To demonstrate its applicability, a synthetic case study is performed without involving the expert elicitation step, integrating real data to model the probabilistic impacts of four risk factors: Production Disruption, Feedstock Availability, Stubble Burning, and Transportation Disruption. The approach provides a robust decision-support tool for identifying critical vulnerabilities and designing targeted solutions for India’s Bio-CNG supply chain.
Download
Paper Nr: 59
Title:

Whether ESG Disclosure Benefits ESG Investment: Empirical Evidence from Chinese Public Firms and Further Opportunities
Authors:

Bingchen Shan, George Q. Huang and Yong-Hong Kuo
Abstract: In recent years, environmental, social, and governance (ESG) is becoming increasingly important as a decision factor in investment boardrooms. A popular trend is to invest in public firms with good ESG disclosure, compared to their counterparts. However, do investors make profits on these ESG investment, especially in Asian market? This study seeks to bridge this gap by investigating all the active equities of Hong Kong H-shares and the Chinese A-shares markets. By utilizing a green factor’s construction method and time series regression, two key results are obtained: 1) first, the good ESG-disclosing firms significantly outperform their peers in Hong Kong market after the 2016’s mandatory disclosure rule is implemented, where ESG fund demand and quarterly earnings are significant drivers for this green outperformance; 2) second, we have derived and compared the average green factor sensitivities across 10 different industries in H-shares and A-shares markets. In accordance with the results, we conclude that an investor-driven incentive to encourage public firms’ ESG disclosure also benefits investors. Furthermore, we claim that this investor-driven ESG disclosure incentive would lead to win-win improvements to different extent across industries. Our research raises further ESG implications for policymakers and public firms.
Download
Paper Nr: 98
Title:

Microgrid Optima: Contract-Aware Optimization and Operational Control of Microgrids with Energy Storage and Renewables
Authors:

Alexander Brodsky and Xu Han
Abstract: Commercial and industrial microgrids increasingly deploy renewable generation and battery energy storage to reduce electricity costs, yet their economic performance is dominated by complex electricity contracts in which supply and distribution demand charges are governed by rolling peak-demand ratchets. Existing optimization approaches often rely on simplified tariff models, short horizons, or heuristic penalties, limiting their ability to capture long-term cost impacts and to support scalable decision making. This paper presents Microgrid Optima, a contract-aware optimization framework that couples billing-period peak-bound synthesis with interval-level operational control. At the billing-period level, we formulate an optimization model that computes optimal supply-demand and distribution-demand bounds for a monthly cycle by directly minimizing contract-defined demand charges. The model explicitly captures contract-specific peak rules, battery power and energy constraints, solar availability, and interval-level demand satisfaction, yielding physically feasible peak bounds that can be enforced without load shedding. At the operational level, we introduce a deterministic, interval-level control algorithm that enforces the optimized peak bounds in real time. Given predicted net demand, the applicable contract-specific bound, and the battery state, the controller computes closed-form battery charge or discharge actions at 15–30 minute resolution, resorting to load shedding only when storage limits are insufficient. By separating peak-bound optimization from high-resolution dispatch, Microgrid Optima enables scalable analysis over full billing periods while preserving operational realism. The resulting demand bounds provide compact, high-fidelity summaries of microgrid flexibility that integrate naturally with long-horizon planning and decision guidance under real commercial electricity tariffs.
Download
Paper Nr: 108
Title:

Intelligent Production Scheduling under Constraints Using LLMs for Industry 5.0
Authors:

Wiem Abbes, Cherifa Nakkach and Yvan Picaud
Abstract: Industry 5.0 manufacturing requires scheduling methods that remain flexible, resilient, and resource-efficient under dynamic constraints. variability Raw-material availability and frequent disruptions, in particular, invalidate fixed-parameter models, degrading solutions from classical optimization and heuristics. This paper presents an end-to-end Large Language Model (LLM) framework for material-constrained production scheduling that reasons directly over heterogeneous inputs-structured production orders, inventory data, and textual operational constraints-and, when available, 3D-derived material estimates. Hard scheduling rules are embedded in the LLM’s reasoning to generate feasible, interpretable schedules without external solvers, while a closed-loop architecture enables real-time monitoring and dynamic re-planning in response to execution deviations and supply variations. We evaluate several contemporary LLMs using semantic similarity metrics based on BERTScore. The best performance is achieved by Mistral Large, with a BERTScore Precision of 0.95, indicating high alignment with reference schedules. These results highlight the promise of LLMs as intelligent planning agents for adaptive, robust, and sustainable scheduling in Industry 5.0 environments.
Download
Paper Nr: 109
Title:

Rubber Belt Production Scheduling by Stochastic Search Leveraging Problem-Specific Characteristics
Authors:

Shunsei Kimura, Hitoshi Iima, Masaki Fujinami and Kazuki Toyomura
Abstract: This study addresses a real-world hybrid flow shop scheduling problem in rubber belt production subject to strict resource constraints. To tackle the difficulty of obtaining feasible solutions, we propose a local search method that leverages problem-specific characteristics. The method prioritizes critical jobs to generate a feasible initial schedule and employs a priority-based strategy during the search to balance feasibility and optimization. Experiments with real data demonstrate that the proposed method effectively reduces constraint violations and improves the objective function compared to manual planning and simple local search.
Download
Paper Nr: 112
Title:

Trust-Building Security Communication for Onboarding into Data Trustee Platforms
Authors:

Jessica Chwalek, Denis Feth and Arghavan Hosseinzadeh
Abstract: Organizations joining inter-organizational data platforms rarely fail due to missing security or privacy safeguards; rather, because those safeguards are not communicated in a way that is comprehensible, decision-ready, and tied to recognizable assurances of legality and accountability. During onboarding, complex legal and technical arrangements must be understood through a small set of decision-critical questions concerning data use, access rights, accountability, and enforceability. This position paper proposes a communication framework that places security and privacy at the center of the onboarding narrative, using plain language, regulatory grounding, and verifiable assurances to support informed decision-making. A pre-survey with prospective data-trust customers indicates that concerns about third-party access (44%) and a preference for supported onboarding formats such as training or webinars (47%) strongly influence participation decisions. Although the empirical input originates from the logistics domain, the proposed approach is domain-agnostic. Its effectiveness is evaluated through a workshop-based assessment focusing on clarity, evidence discoverability, decision efficiency, and reduced clarification effort.
Download
Paper Nr: 11
Title:

Simulation-Based Energy Efficiency Analysis of Lighting Systems in Outpatient Hospital Areas: A Case Study
Authors:

Ruth Ruiz Diaz and Eduardo Redondo
Abstract: The growing demand for energy-efficient systems in healthcare highlights the importance of optimized lighting solutions. This study evaluates the current lighting systems and proposes alternatives for three outpatient areas in a public hospital in Paraguay: the waiting room, ultrasound room, and tomography room. A diagnostic phase, based on EN 12464-1 standards, identified deficiencies such as low illuminance levels and poor uniformity. Using DIALux Evo simulation software, customized LED-based configurations were designed to improve performance while ensuring compliance with international lighting standards. A technical-financial analysis assessed the feasibility of each alternative in terms of implementation, operation, and maintenance. While initial and operational costs were higher than the baseline, the proposed designs offer significant improvements in energy efficiency, lighting quality, and safety. The study demonstrates how simulation-based approaches, combined with economic analysis, can support sustainable lighting upgrades in healthcare environments. The proposed framework is replicable and adaptable to similar facilities in developing regions.
Download
Paper Nr: 54
Title:

A Predator–Prey MOEA/D with Deep Q-Network for Scheduling Problem
Authors:

Maha Ben Hamida, Ameni Azzouz and Lamjed Ben Said
Abstract: The predator-prey mechanism introduces adaptive evolutionary pressure into the core of the framework for the Green Flexible Job Shop Scheduling (Green FJSP) problem. This problem involves conflicting objectives, such as minimizing production time and reducing energy consumption, making it difficult to solve. In the proposed hybrid approach, predators accelerate the search for high-quality solutions, while prey maintain population diversity and prevent early convergence. To support this process, a Deep Q-Network (DQN) is integrated into the MOEA/D-PBI algorithm to guide operator selection, enabling flexible control of exploration and exploitation. The combination of predator-prey dynamics and DQN-based learning creates a self-adaptive balance during the optimization process, improving both convergence speed and solution quality. Experimental results on Green FJSP benchmarks show that our proposed approach achieves consistently better performance compared to the reference methods, offering a better trade-off between production efficiency and energy sustainability.
Download
Paper Nr: 79
Title:

Soft Computing-Based Decision Support for Sustainable Waste Collection Operations
Authors:

Jean P. Morán-Zabala, Juan M. Cogollo-Flórez and Alexander A. Correa-Espinal
Abstract: The growing complexity of waste collection (WC) systems necessitates analytical frameworks that integrate expert knowledge with real−world operational data under uncertainty. This study proposes a soft computing– based decision support approach using Fuzzy Cognitive Maps (FCMs) to model the causal structure of WC operations in a medium-sized Colombian city. The framework captures non-linear interactions among key operational variables, including collected tonnage, travelled distance, number of trips, compaction frequency, fuel consumption, and total collection time. The methodology combines correlation-based structural identification with iterative causal inference to analyze system dynamics and convergence. Results show rapid model stabilization and a consistent hierarchy of causal influences, with workload intensity and total collection time as the dominant drivers, followed by distance, trip frequency, and fuel consumption. The proposed operational specification limits are empirically validated and shown to be consistent with observed process behaviour. These findings demonstrate the ability of FCMs to enhance interpretability, support adaptive decision-making, and identify leverage points for sustainable route planning and resource optimization, providing a scalable and empirically grounded complement to classical optimization and simulation approaches in urban waste management.
Download
Paper Nr: 88
Title:

Experiential Learning Meets Generative AI: Teaching Quantitative Methods in Systems Engineering
Authors:

Jairo R. Montoya-Torres
Abstract: Generative artificial intelligence (GenAI) is rapidly changing how quantitative methods are taught and learned in higher education, by personalizing feedback and supporting modelling and problem-solving skills. However, educators still face challenges in integrating these technologies into their courses. Taking the case of a quantitative course at the master level in Supply Chain Analytics, this paper introduces a conceptual framework, which integrates experiential learning with GenAI to help students develop deeper understanding and critical thinking for optimization courses. The framework is however intended to be generalized. Also, the paper illustrates how the framework can be implemented, showing practical strategies for using experiential activities and GenAI support. The goal is to position the coupling of Experiential Learning and GenAI tools as a collaborative partner in the classroom, without replacing human interactions.
Download
Paper Nr: 100
Title:

Multi-Objective Multi-Factory Scheduling with Robust Optimization for Inter-Factory Routing
Authors:

Yurika Suzuki and Sumika Arima
Abstract: This study addresses robust multi-factory management under dynamic fluctuations in production capacity. As manufacturers expand multi-factory operations to meet market needs and ensure stable supply, semiconductor firms increasingly add new production buildings within the same area, making inter-building transfers common. However, scheduling methods that account for such transfers-and the capacity variations that cause them-remain insufficient. To resolve this issue, the study proposes a multi-factory scheduling method that incorporates transfers triggered by factory or equipment downtime. Building on the P3D-QAP2MF algorithm, which optimizes cross-factory load assignment for Q-time compliance, due-date adherence, and setup reduction, the proposed method adds a new objective: minimizing transfer time through continuous in-factory processing. Numerical experiments using a five-building model show substantial improvements over Company A’s Earliest Start Machine method, notably in setup rates, cycle time, and transfer time, with slight gains in due-date performance. In particular, these results demonstrate a significant advantage for semiconductor manufacturing, where small performance gains have large economic impacts (US$1 to US$10 million by 1%).
Download