This study develops a multi-scenario coordinated scheduling framework for green port-integrated energy systems that extends capacity-based planning by explicitly incorporating renewable output variability. The study examines how renewable energy integration costs in port environments are affected by output variability, in addition to average generation levels. Based on a mixed-integer linear programming model, the framework integrates conventional generators, renewable sources, energy storage systems, and grid interactions. A scenario taxonomy is established across load intensity, solar irradiance, and wind speed to evaluate multiple scenario combinations. Five experiments investigate baseline operational economics, renewable variability impacts, load intensity effects, carbon pricing mechanisms, and energy storage coordination. The results indicate that renewable energy integration costs are affected more by output variability than by average generation levels; that the relative importance of the economic and reliability functions of energy storage changes as load intensity increases; and that storage performance depends on matching capacity-to-power ratios to port-specific load fluctuation patterns. The study decomposes the value of energy storage into economic value from peak-valley arbitrage and reliability value from avoided load shedding. It also develops a load–storage configuration principle showing that performance depends on aligning capacity-to-power ratios with port-specific load dynamics rather than increasing capacity alone. The findings suggest that ports should consider both the quantity and variability of energy supply and use storage to balance cost, reliability, and environmental objectives under changing market conditions.
Digital logistics environments increasingly require human-centric training approaches that support procedural reliability and operational performance. Virtual Reality (VR) has been adopted for immersive training in such contexts, yet evidence regarding its relative effectiveness compared with conventional instructional methods remains mixed. This study presents a controlled experimental comparison of VR-based and text-based instruction for manual order picking tasks, integrating objective performance measures with structured self-reported evaluations. Results indicated no statistically significant differences between training conditions in task completion time or error rates. Participants in the VR condition reported statistically higher ratings of confidence, perceived effectiveness, and perceived applicability to real-world operational settings. At a ten-day follow-up, perceived knowledge retention remained high and comparable across conditions, while expectations of future performance improvement were statistically higher in the VR group. This pattern suggests that the benefits of VR may have been expressed more strongly through perceived engagement, spatial awareness, and user experience than through immediate objective performance differences. VR may be positioned as a complementary instructional modality associated with higher perceived competence and readiness within digital intralogistics environments, rather than as a replacement for conventional instruction. Despite limitations related to sample composition and reliance on self-reported measures, the study provides controlled empirical evidence on instructional modality differences in manual order picking training.
In automated container terminals (ACTs) utilizing automated lifting vehicles (ALVs), the apron and block buffer zones are critical components that decouple the operations of the three key types of handling equipment: quay cranes (QCs), ALVs, and automated stacking cranes (ASCs). By temporarily storing containers, these buffer zones allow for asynchronous operations among equipment, significantly enhancing equipment utilization and system flexibility. However, their limited capacities also create a system bottleneck, and inefficient scheduling can easily lead to mutual equipment waiting and operational blocking. To address this challenge, this paper investigates the integrated scheduling problem of QCs, ALVs, and ASCs with limited apron and block buffer zone capacities. The key decisions involve the ASC-job sequence, job-to-ALV assignment, and ALV-job sequence. First, a mixed-integer linear programming (MILP) model is formulated using different approaches to model the capacity constraints of the distinct buffer zones. Second, an encoding method for key decisions is developed to generate the insertion sequence of container jobs. Then, by analyzing the characteristics of buffer zones, two types of sequence rules are designed, and a method for calculating terminal equipment schedules is derived from the proposed model. Finally, a sequential insertion algorithm (SIA), an SIA-based adaptive large neighborhood search algorithm, and a lower bound method are proposed to solve mid- and large-sized cases. Numerical experiments demonstrate that the proposed algorithms perform well and offer advantages over off-the-shelf solvers. Based on the experimental findings, managerial implications for the operation and management of automated container terminals are discussed.
Additive manufacturing delivers unique advantages over conventional methods, but efficient operation of large-scale 3D printing farms remains a challenge. This paper studies the integrated nesting and scheduling problem for non-identical parallel printers under calendar constraints that restrict batch start times to predefined operational shifts. We introduce a mixed-integer linear programming model capturing setup dependencies that require human intervention and continuous, unattended printing. The objective is to minimize the makespan by jointly optimizing part nesting and shift-aligned start times, thereby maximizing machine utilization around the clock. To address large instances, we propose a two-stage solution: a fast constructive heuristic generates an initial schedule, which is then refined by an adaptive large neighborhood search (ALNS). Computational results on problems solved optimally by the model show that the heuristic produces high-quality initial solutions in negligible time, while the ALNS achieves near-optimal schedules with an average deviation of 0.18
The Discrete Time/Resource Trade-Off Problem (DTRTP) under uncertainty poses a significant challenge in project management, as it requires simultaneously optimizing scheduling efficiency and robustness against unforeseen disruptions. To address this challenge, this study proposes a hybrid metaheuristic framework integrating the Starting Time Criticality (STC) concept for the systematic generation of robust project schedules. The core contribution lies in formulating an STC-based robust scheduling model, in which activity work content is used as a surrogate measure to capture the impact of uncertainty on schedule robustness. Based on this model, a hybrid metaheuristic framework is developed, embedding STC into solution evaluation and search guidance while remaining flexible to different search engines. Representative implementations using Genetic Algorithm (GA) and Differential Evolution (DE) are constructed to demonstrate the framework’s effectiveness and generality. Extensive computational experiments show that the proposed STC-based framework significantly outperforms existing state-of-the-art robust scheduling models in terms of four performance measures. Overall, the proposed approach provides a practical and extensible methodology for enhancing schedule robustness and resilience in complex project environments.
We consider a scheduling problem for two-stage flexible flow shops with parallel batch processing machines (BPMs), motivated by settings found in the diffusion area in semiconductor wafer fabrication facilities. An objective function blending the total weighted tardiness (TWT) and the total electricity cost (TEC) is used under a time-of-use (TOU) tariff. Initial maximum time lags, i.e. time constraints for the operations on the first stage and regular time constraints between the operations on the first and second stage are taken into account to prevent native oxidation and contamination effects on the wafer surface. Problems with time constraints have not been discussed in the literature so far for energy-aware scheduling. A mixed-integer linear programming (MILP) formulation is established for this scheduling problem. Moreover, a genetic programming (GP) procedure is designed to automatically discover priority indices within a new heuristic scheduling framework that is able to avoid time constraint violations. The framework is based on a lead time iteration approach to set internal due dates for the first stage and release dates for the second stage and a time window decomposition (TWD) scheme together with a decision theoretic heuristic (DTH). The GP approach uses specific terminals to avoid the violation of time constraints. Moreover, violations are penalized in the GP approach. Results of designed computational experiments are reported that demonstrate that the learned priority indices lead to high-quality schedules with only a very small number of time constraint violations in a short amount of computing time. Tradeoffs between the TWT and TEC values of the resulting schedules can be obtained by the proposed approach.
With the continuous development of intelligent manufacturing, automated guided vehicles (AGVs) have become a crucial component of transportation resources within workshops. Consequently, explicitly incorporating AGV transportation time into workshop scheduling is of considerable practical importance. This paper investigates the flexible job-shop scheduling problem with AGV transportation time (FJSP-AGV). First, a hierarchical action-space structure is proposed to address the multi-action decision-making challenge inherent in the FJSP-AGV. Second, an integer programming model is established for the FJSP-AGV, aiming to minimize the makespan. Subsequently, FJSP-AGV is modeled as a multi-agent Markov decision process (MMDP), in which a combination of a graph isomorphism network (GIN) and a multilayer perceptron is employed to encode and decode the state information of operations, machines, and AGVs. The Proximal Policy Optimization (PPO) algorithm is used to optimize the performance of the decision model. Finally, the trained model is evaluated on both benchmarks and randomly generated large-scale instances. For standard benchmark instances, the proposed approach achieves a 7
In smart semiconductor manufacturing, wafer metrology plays an important role in ensuring production quality and efficiency. Extensive production metrology however implicates substantial costs in terms of equipment and time needed to perform the measurements. Achieving efficient metrology plans is therefore highly desirable. When the characteristics being monitored are spatially correlated it is possible to develop dynamic sampling strategies that exploit this correlation to reduce the number of measurement sites needed for full process visibility. These approaches typically employ historical data to identify the subset of locations to measure for each wafer, and statistical models that can reconstruct the complete data from the reduced set of measurements. Existing dynamic sampling strategies have focused on trading off reconstruction accuracy with the number of measurements points, without considering constraints on the temporal horizon, that is, the number of wafers that need to be processed before all available locations are measured at least once. This is an important parameter in practical applications as it defines the maximum number of wafers for which previously unseen abnormal behavior can go undetected. In this paper we formulate the temporally bounded dynamical sampling problem and explore a number of different strategies for solving it. These strategies focus on the best way to select new sites to visit for each wafer in order to comply with the temporal constraint, while minimizing the impact on process visibility. The effectiveness of the proposed methods is tested using both simulated and real world semiconductor manufacturing case studies. Results show that an algorithm, denoted DFSCA-Greedy, that distributes the number of forced sites evenly across the measurement plan, and selects the sites with a variant of the FSCA optimum site selection algorithm, achieves the best overall performance.
In practical assembly line balancing problems, obtaining complete information on the probability distribution of task time is extremely challenging. Often, only partial information such as means, variances, and medians is available, making it difficult to determine the specific distribution type. To address this issue, this article proposes a distributionally robust optimization method for assembly line balancing under conditions of distributional uncertainty. The method establishes uncertainty sets based on mean and covariance matrices, accounting for both known and unknown matrices. A distributionally robust model is then formulated and solved using an improved genetic algorithm and simulated annealing algorithm separately. The results demonstrate that the improved genetic algorithm performs better in solving the distributionally robust assembly line balancing problem, particularly when the given completion probability is low, resulting in fewer workstations.
Driven by the dual-carbon strategy, green innovation, as a kernel element of the new quality productivity, is reconfiguring the path of high-quality development of the manufacturing industry. In view of the rigid requirements of green environmental protection, it is difficult for the manufacturing industry to realize green transformation only by itself, and the synergy of multiple subjects has become the key. In contrast to the limitations of traditional research focusing only on the two-subject perspective, this paper takes the unique perspective of synergy of multiple heterogeneous subjects in the green innovation network, incorporates the government's environmental regulation, manufacturing enterprises' production mode, and consumer's green preference into the research framework of the synergy mechanism of the green innovation network, constructs the three-party evolution game model of the government, manufacturing industry, and consumers, and simulates the effect of changes in the parameters of the network of heterogeneous subjects on the evolution of the synergy by using Matlab. evolution of the heterogeneous subjects. The study finds that the collaborative evolution of green innovation network subjects shows significant strategy dependence, in which the government regulation forms the core anchor of the evolutionary stabilization strategy through the dual mechanism of incentives and constraints; the consumers' willingness to pay for low carbon is heterogeneous depending on the type and intensity of the government's environmental regulation tools and the interaction between the manufacturing behaviors of the enterprises. This study innovatively reveals the dynamic coupling law of the behavioral strategies of heterogeneous subjects, and purposefully proposes the synergistic strategy of green innovation network of multiple heterogeneous subjects. Its theoretical contributions include: (1) constructing a dynamic evolution analysis framework of “institutional design-market response-technological leap”; (2) identifying the critical conditions for government-manufacturing-consumer synergy in the green innovation network synergy mechanism. The research results provide a new analytical paradigm for solving the “island effect” of green innovation in the manufacturing industry, and have decision-making reference value for improving the green innovation network governance system.
This study investigates energy-aware production control in a make-to-stock manufacturing system with two serially connected, non-identical machines operating in multiple energy modes such as production (on), standby, and off. Motivated by the growing need to reduce industrial energy consumption while maintaining service levels, we model the system as a continuous-time Markov decision process (MDP) and solve it exactly using a linear programming approach. Our analysis explores how demand rate, sales price, production capacity, and energy costs interact with the bottleneck location to influence key performance metrics such as long-run average profit, inventory levels, service levels, and machine utilization. We further compare the optimal energy-efficient policy with the widely used always-on policy, under which machines remain powered on and operate in either Production or Standby mode to mitigate lost sales. Numerical experiments based on a full factorial design reveal that system profitability and efficiency are primarily driven by demand intensity, sales margins, and bottleneck shifts, whereas variations in energy tariffs have relatively minor effects. Importantly, we show that the always-on policy can approximate the optimal solution under high demand or high-margin conditions, but leads to substantial inefficiencies when demand is low. These insights provide practical guidance for implementing energy-saving strategies in industrial production lines without compromising service performance.
This study addresses the mixed-case multi-size palletization problem (MC-MSPP) faced during humanitarian logistics. The problem includes effective loading of mixed items onto heterogeneous pallets in time and secure delivery. The complexity of the problem lies in optimizing space utilization while minimizing the number of pallets. The complexity of task compounded by the diverse characteristics of items and pallet dimensions. To address this, we develop a mathematical formulation that integrates critical constraints on orthogonal item rotation, stability, fragility, compressibility, and incompatibility. This ensures safe and effective stacking under diverse pallet weight and volume limitations. Our solution approach consists of discrete integer optimization and a three-stage layer-based constructive heuristic approach. The heuristic effectiveness is validated against an exact approach with novel datasets and established benchmarks. The significance of the research lies in its ability to bridge the gap in the practical implementation of mixed case packing scenarios with high-quality solutions within 2 min—making it ideal for real-time logistics operations.
In response to the rapid growth of e-commerce, fulfillment centers face the challenge of managing an increasing volume of online orders. Unlike traditional logistics, e-commerce involves frequent and smaller orders that require packaging in individual parcel boxes. To simplify the packaging process, e-commerce companies have standardized the sizes of parcel boxes, necessitating the addition of void fills to protect small or fragile items. However, the irregular placement of differently sized items within parcel boxes requires considerable manual effort for void fill insertion. Consequently, automated operations for void fill insertion at the stage of the packaging process in e-commerce fulfillment centers are required. Thus, we propose a comprehensive process for void fill insertion into the parcel boxes, from pre-processing 3D images of the items to planning the route of the void fill insertion robot. To determine the robot’s route, we propose a mathematical optimization model. Due to the model’s inherent complexity, we also develop an efficient heuristic algorithm that offers computational time efficiency. The results of numerical experiments demonstrate that our proposed methodology enhances the efficiency of void fill insertion by resolving issues within a short computation time. Additionally, we provide valuable managerial insights which can help improve the operations in the e-commerce fulfillment centers.
The flexible job-shop scheduling problem (FJSP) is a critical domain within production planning and control. It has garnered significant interest due to its capacity to optimize resource allocation, enhance production efficiency, reduce manufacturing costs, and bolster the coordination of the entire production process. As manufacturing systems evolve to incorporate advanced technologies, the integration of automated guided vehicles into these systems introduces a new layer of complexity to scheduling. The FJSP with automated guided vehicles (FJSP-AGV) considers the transportation time of AGVs and adds the complexity of AGV allocation to the FJSP, significantly increasing the uncertainty and complexity of the scheduling process. This paper proposes a scheduling optimization algorithm based on a heterogeneous graph neural network that integrates multiple attention mechanism, designed within an end-to-end learning framework to achieve scheduling of FJSP-AGV. First, a heterogeneous graph model of the workshop state is construct to extract the complex relationships among operations, machines, and AGVs. Next, we utilize multiple attention mechanism to process the state features. Subsequently, we optimize the decision model's performance using the proximal policy optimization (PPO) algorithm. Finally, the algorithm framework is trained using randomly generated cases, and the trained model is combined with Monte Carlo Tree Search (MCTS) and applied to both standard cases and randomly generated cases. The results demonstrate that, in standard cases, our algorithm outperforms existing mainstream reinforcement learning algorithms, while also exhibiting superior generalization in randomly generated cases.
In the era of Industry 4.0, collaborative robots (cobots) have emerged as pivotal technologies reshaping the manufacturing landscape. The collaboration between humans and cobots brings higher productivity and flexibility. Specifically, cobots mitigate physical overload by assuming repetitive and labor-intensive tasks, thereby improving worker safety and reducing fatigue. This study attempts to quantify ergonomics in human–robot collaborative two-sided assembly line balancing problems (HRCTALBP) through a fatigue-and-recovery criterion. Furthermore, in light of global energy sustainability concerns, cobot energy consumption is integrated as a critical optimization parameter. To address these challenges, a multi-objective framework is proposed that simultaneously minimizes cycle time, ergonomic risk, and energy consumption. A novel mixed-integer programming model is formulated, incorporating multi-skilled workers and adaptive cobot capabilities. In addition, an improved multi-objective migrating birds optimization algorithm is developed, featuring a multi-population co-evolution mechanism to enhance solution diversity and convergence. The algorithm’s efficacy is rigorously validated against three different algorithms through comparative experiments. Finally, a real-world case study from the automotive industry further demonstrates the practicality of the proposed approach. Results highlight HRC as an ergonomic and efficient solution, while the derived Pareto-optimal solutions offer actionable insights for production managers in configuring assembly lines that balance productivity, worker well-being, and sustainability.
This paper studies a photolithography scheduling problem with dynamic job arrivals, uncertain setup and processing times, and random machine breakdowns with stochastic downtimes. To address this real-life motivated complex problem, we adopt a rolling horizon approach, which involves solving a static and deterministic scheduling problem, executing a part of the proposed schedule, and rescheduling after a predetermined interval to create an updated schedule. To solve the static problem, a new constraint programming (CP) formulation is introduced. A case study is performed at a global semiconductor manufacturer in which a data-driven simulation model is used to accurately mimic the dynamics of the real-world photolithography area. In this simulation model, our proposed CP formulation is deployed with the rolling horizon approach and benchmarked against an alternative CP formulation from the recent literature, some well-known dispatching heuristics, and the current practice.
As the use of autonomous mobile robots (AMRs) has increased in material handling systems, it has become essential to incorporate their unique characteristics and constraints into the pickup and delivery problem. Additionally, request scheduling and charging arrangements should be considered simultaneously to maintain battery feasibility during transportation tasks. Realistic charging and discharging assumptions must also be integrated into the model. Minimizing request delays is crucial, as delays can negatively impact productivity. This study addresses the scheduling problem in material handling systems, aiming to minimize the total tardiness of transportation requests while considering the specific characteristics of AMRs. This study also considers the partial recharging policy, nonlinear charging, and load-dependent discharging. A mixed-integer linear programming model was developed and an adaptive large neighborhood search (ALNS) algorithm was constructed to minimize total tardiness, with travel time minimization also included. A full factorial design with seven factors was conducted, revealing that layout, transportation request size, the number of AMRs, and battery levels are significant factors influencing total tardiness, while all factors significantly affect travel time. The results underscore the importance of considering charging and discharging mechanisms in effectively solving this problem.
Semiconductor manufacturing lines are constantly evolving, with new technologies and products being introduced at an accelerated pace in the market. In order to keep up with this rapid pace, it is essential to expedite the ramp-up of the production line, as delays can result in missed market opportunities and lost revenue. Expedited ramp-up requires careful planning and execution from recipe setup on the target machines to the final qualification for the mass production. Especially, scheduling engineering lots to collect data for the qualification process amid the current production schedule has been an important and difficult problem to resolve. However, it is currently managed mainly by the engineers’ experience and knowledge, which has led to variability and inefficiency in the ramp-up process. As a consequence, Samsung Electronics started to resolve this issue and introduced a systemized approach using a logic-based system and discrete-event simulation, which helps prioritize engineering lots in harmony with the current production schedule. In this work, we introduce the concept of our approach and show its performance through an actual field test.
Human–robot collaboration (HRC) offers new avenues for balancing efficiency and well-being in modern assembly systems. This study develops a Mixed-Integer Linear Programming (MILP) formulation for the Mixed-Model Assembly Line Balancing Problem (MMALBP) that explicitly embeds multi-dimensional ergonomic considerations into the balancing process. The proposed model, termed the Ergonomics-Oriented Mixed-Model Assembly Line Balancing Problem with Human–Robot Collaboration (Ergo-MMALBP-HRC), is formulated as a multi-objective optimization framework. It simultaneously minimizes cycle time, overall ergonomic risk (ER), and ergonomic category (EC) scores by integrating the physical, environmental, and psychosocial dimensions of ergonomics. Ergonomic risks are assessed using a multi-method framework and synthesized using a fuzzy logic–based approach to provide a holistic representation of task-level exposures. A computational analysis based on benchmark-derived configurations demonstrated the model’s effectiveness, showing that the inclusion of ergonomic constraints considerably mitigates risks without compromising productivity. On average, EC and ER decreased by 27
With the rise of mass individualized customization, the increase in customer-specific requirement data has significantly complicated product configuration design. Traditional configuration methods often simplify the complexity of configuration contents, but this approach fails to deliver solutions that fully meet customer requirements. To address this, this paper proposes an intelligent guided configuration design method based on a knowledge graph. The method begins by constructing a multi-domain product knowledge graph from enterprise design resources. Knowledge fusion across domains is then achieved using operators from C-K theory, which facilitate conceptual expansion and innovative reasoning. Through interactive mapping between customer requirements and the knowledge graph, customized requirement nodes and components are identified to generate configuration solutions. The proposed method demonstrates superior capability in handling complex, multi-domain knowledge and dynamic requirement reasoning compared to traditional approaches, as preliminarily validated through a case study on air purifier configuration.
