Scholarly Works, Industrial and Systems Engineering

Permanent URI for this collection

https://hdl.handle.net/10919/24292

Research articles, presentations, and other scholarship

Browse

Now showing 1 - 20 of 594

Efficiency and Equity in Screening for Participants in Representative Cybersecurity Studies
Chen, Po-Yu; Oh, Jinwoo; Hsing, Hsiang-Wen; Lau, Nathan; Beltz, Brandon; Wu, Peggy; Zhu, Quanyan (Springer, 2026-07-31)
Human factors research in cybersecurity confronts the practical challenge of how to recruit and screen participants with the skills necessary to perform demanding, ecologically valid tasks. Credible results from hacking studies rely on participants possessing sufficient technical skills to maneuver inside a representative network in a cyber range. Screening via self-reports can be unreliable, while rigorous skills testing is resource intensive. This paper presents empirical research investigating the effectiveness of cybersecurity certifications and cyber range skills tests for screening domain-specific competence. The Guarding Against Malicious Biased Threats (GAMBiT) project recruited participants with a two-stage screening process. The project screened 269 candidates for a final sample of 61 qualified participants. We compared the capability between the Offensive Security Certified Professional (OSCP) certification and a custom Capture-the-Flag (CTF) assessment (“Cedar Bunny”) at predicting actual performance in a two-day cyber range experiment. Results indicate that OSCP holders scored significantly higher on the skills test (p < 0.001) and progressed further in the hacking experiment (p = 0.002) than non-holders. Skills test scores correlated moderately and positively with hacking progress (ρ = 0.344). Furthermore, hierarchical regression analysis demonstrated that the skills test provided incremental predictive validity (ΔR2 = 5.2%) beyond certification alone. These findings suggest that while certifications are efficient proxies and skills tests can be effective supplements for identifying viable, albeit non-certified, participants. We propose a mixed screening approach to balance recruitment efficiency and participant diversity.
Evaluating model-estimated shoulder muscle activity during overhead work with varied task demands and exoskeleton use
Li, Lingyu; Behjati Ashtiani, Mohamad; Kim, Sunwook; Nussbaum, Maury A. (2026-05-21)
Passive arm-support exoskeletons (ASEs) can reduce shoulder stress during overhead work. However, the effects of ASEs are task- and device-specific, and current human-subjects evaluation protocols are time-consuming and resource-intensive. Musculoskeletal modeling could simplify ASE evaluation, such as by estimating muscle activation in place of measuring electromyographic (EMG) data. We assessed shoulder muscle activities estimated using a commercial biomechanical modeling system during dynamic overhead push tasks at different heights and force directions, performed with/without an ASE. We compared model estimates to normalized EMG using pattern similarity and magnitude difference metrics. Overall, model performance was good at lower task heights and declined at higher heights; ASE use further reduced performance, especially at higher task heights. These findings suggest that model-estimated shoulder muscle activity may be reasonably accurate under specific task conditions, but that improvements to musculoskeletal models are needed to make these models suitable for a broader range of tasks.
Evaluating the accuracy and feasibility of a commercial AI-powered ergonomic assessment system for automotive assembly work
Zahabi, Saman Jamshid Nezhad; Kim, Sunwook; Nussbaum, Maury A.; Porto, Ryan; Lim, Sol (Elsevier, 2026-10-01)
Accurate ergonomic risk assessment is essential to prevent work-related musculoskeletal disorders. Artificial intelligence (AI) and computer vision offer new opportunities for automatic risk assessment, but accuracy and usability remain uncertain. We evaluated a commercial, AI-powered, ergonomic assessment system through two studies: In an exploratory evaluation (Study 1; n = 10), we compared risk estimates and joint angles using three camera placements vs. a gold-standard 3D motion capture system. In Study 2 ( n = 10), we interviewed ergonomics professionals who used the software system for automotive assembly tasks. We found substantial inaccuracies in joint angles (especially left elbow and bilateral wrists) and risk estimates, and important variability across camera setups. Although the software system was reported as easy to use and visually intuitive, concerns were noted about reliability—particularly in cluttered environments and tasks involving fine hand motions. While promising, these systems may require further refinements to improve accuracy and adaptability.
Oxalate-Linked Electrochemistry Enables Early Detection of Sclerotinia Blight in Peanut Crops Using 3D-Printed Nanostructured Pt Sensors
Erukainure, Frank Efe; Johnson, Blake N.; Langston, David B.; Abhilash, Chandel K. (ACS, 2026)
Sclerotinia minor (Sclerotinia blight) is a devastating pathogen in peanut production, with severe outbreaks causing up to 50% yield loss due to rapid oxalic acid (OA) accumulation. Early diagnosis is challenging because canopy-level symptoms typically emerge only after infection is well established, while early cues are subtle, stem-localized, and nonspecific; in contrast, molecular assays are time- and resource-intensive. Despite established mechanistic links between oxalate accumulation and disease progression, so far, there are no sensors developed or tested for detecting Sclerotinia blight in peanut plants. This paper reports a low-cost, lithography-free and label-free electrochemical sensor for metabolite-targeted, pre-symptomatic monitoring of S. minor in peanut plants based on clear mechanistic links between oxalate accumulation and disease progression. The sensor platform comprises 3D-printed resin substrates with Platinum (Pt) electrodes and a nanostructured reduced graphene oxide (rGO)–chitosan interface functionalized with an oxaloacetic acid (OAA) interfacial layer. Using ferri/ferrocyanide as a redox probe, the sensor exhibited a linear calibration to oxalate (prepared from OA) from 0.05 μM to 1 mM (R2 = 0.99), with a sensitivity of 6.37 μA/decade, limit of detection of 17.6 nM, and excellent coefficient of variation of 0.93–3.32% across standards (n=4). In real plant trials, stem sap from S. minor–inoculated peanut plants produced significantly elevated voltammetric responses relative to healthy and Nothopassalora personata controls as early as 5 days post-inoculation (dpi), enabling longitudinal monitoring through 20 dpi (p < 0.001). Oxalate-equivalent mapping showed progressive increases in infected plants, reaching mM levels by 20 dpi, while the validated sensor readings agreed with those of a commercial assay during oxalate-linked S. minor detection in peanut plants. To our knowledge, this is the first demonstration of a 3D-printed, Pt-based electrochemical platform validated with peanut plants for early, metabolite-linked detection of Sclerotinia blight, providing a practical foundation for point-of-need surveillance and precision disease management in peanut production systems.
Effectiveness and usability of a trunk posture feedback system: A longitudinal field study for up to six weeks in a distribution center
Choi, Jiwon; Kim, Sunwook; Lim, Sol; Porto, Ryan; Nussbaum, Maury A. (Elsevier, 2026-10-01)
Postural feedback systems are a potential ergonomic intervention to reduce postural exposures and mitigate musculoskeletal disorder risk, yet field-based evidence on their long-term effectiveness remains limited. We investigated a commercially available postural feedback system implemented in a logistics environment, which provided auditory and vibrotactile feedback in response to excessive trunk flexion. Thirty-two workers used the system for up to six weeks. No initial improvements in postural exposures were observed between baseline (Day 1, no feedback) and the first few days of feedback (Days 2–4). Mixed-effects models indicated no sustained improvements over time, with substantial variability across participants. Participants reported that the system was easy to use and increased postural awareness, although some noted inconsistencies in feedback. These findings suggest that more precise and context-aware feedback systems might better promote sustained behavioral change, though our results should be interpreted in the context of the specific system and work environment evaluated.
Wearing an Arm Support Exoskeleton Does Not Affect Balance but May Decrease Dynamic Stability During a Step-Down Maneuver
Arippa, Federico; Barr, Alan; Phillips, Brandon; Kim, Sunwook; Nussbaum, Maury A.; Harris-Adamson, Carisa (SAGE Publications, 2026-04)
Objective: Assess the effects of arm support exoskeletons (ASEs) on dynamic balance during dynamic tasks. Background: ASEs can reduce muscle activation during labor-intensive tasks, potentially alleviating fatigue, discomfort, and injury risk. However, implications on worker safety, particularly regarding altered balance, remain a concern. Methods: We evaluated the effects of three different ASEs on dynamic balance using postural sway parameters and the dynamic postural stability index (DPSI). Twenty-three healthy volunteers (7 F) performed a single-leg, step-down maneuver with and without ASEs. Results: Sway did not differ substantially across conditions, except for a slight increase in sway area in one comparison (18.3%, ηp2≈0.07). In contrast, DPSI increased in all ASE conditions compared to no-ASE (6.1–10.1%, ηp2≈0.05–0.10). A main effect of sex was found for sway and dynamic stability metrics, with females exhibiting greater postural excursions (13.5% higher sway and 9.3% higher DPSI on average, ηp2 up to 0.24). Height and body mass were negatively correlated with sway and DPSI parameters, suggesting a potential role of individual anthropometrics in modulating balance performance. Conclusion: ASEs do not impair balance during dynamic tasks such as a step-down maneuver and single-leg stance, but they may affect stabilization strategies, especially for individuals with low body mass and height. Application: These findings support the potential for a cautious integration of ASEs in industrial settings, as these devices appear to have a minimal impact on balance during moderately dynamic tasks, but observed differences in DPSI highlight the need for careful evaluation in specific populations.
Feasibility of forecasting self-injurious behavior among autistic youth using wearable sensors and machine learning models
Kim, Sunwook; Cantin-Garside, Kristine D.; Nussbaum, Maury A. (Springer Nature, 2026-04)
Self-injurious behavior (SIB) is a substantial clinical challenge for many individuals on the autism spectrum, and support strategies are often only reactive. Forecasting of SIB could enable timely support, especially using wearable sensors, but its feasibility is not well understood. We evaluated SIB forecasting using a previously collected dataset (n = 9) comprising motion and physiological data. We compared the performance of four machine learning models-Random Forest, AdaBoost.M2, Long Short-Term Memory (LSTM), and a Double-Stacked LSTM-across five forecast horizons (3s to 120s) and three feature sets: Motion-Only (from accelerometers), Physiological-Only (e.g., heart rate, skin conductance), and Combined. Performance was measured with a range of metrics, using Leave-One-Subject-Out cross-validation. We found a significant main effect of forecast horizon on the Area Under the Precision-Recall Curve; performance rose from near-chance at short horizons to having median scores above chance at one minute or longer. While aggregated results showed no significant differences between models or feature sets, subject-level analysis suggested predictive feasibility and that the optimal model configuration were highly person-specific. Our findings demonstrate that forecasting using wearable sensor data is feasible, but the substantial performance variability highlights a critical need for person-specific approaches to enable the development of clinically useful, proactive support systems.
An exploration of Kaizen events in hospitals using a systematic literature review and bibliometric analysis
Harry, Kimberly D.; Van Aken, Eileen M.; Glover, Wiljeana J. (Emerald, 2025-05-06)
Purpose: This study aims to comprehensively survey the existing body of knowledge on Kaizen events (KEs) in hospitals, highlighting key dimensions such as the maturity of KE research, trends in publications and authorship, the networks of collaborating research teams and communities and commonly utilized research methods. Design/methodology/approach: To assess the current body of knowledge on this topic, a systematic literature review (SLR), integrated with bibliometric analysis, was conducted, drawing from three major databases: ProQuest, Web of Science (WOS) and Engineering Village. Findings: The SLR process yielded 64 technical papers for analysis. Bibliometric findings reveal that hospital KE research is a pertinent and engaging topic for academic scholars and industry practitioners. While evidence of collaboration exists within the research community, there is significant potential to leverage further collaborative efforts for even greater research advancements and impact. Moreover, the findings also highlight opportunities for further development and advancement in this research area by quantitative approaches. Research limitations/implications: This work was tailored to capture relevant works from a select number of platforms with studies in English. Future research could expand the scope by incorporating additional databases and exploring other key maturity dimensions to achieve a more comprehensive assessment and synthesis. Originality/value: This study provides a holistic and comprehensive evaluation of hospital KEs, utilizing a rigorous systematic approach across three major scholarly databases. Additionally, it aims to advance this critical knowledge area by elaborating on current trends and developments within this research domain.
Vision-language models for occupational physical exposure assessment: Classification and temporal segmentation of manual material handling tasks
Rajabi, Mohammad Sadra; Ojelade, Aanuoluwapo; Kim, Sunwook; Nussbaum, Maury A. (Elsevier, 2026-06)
Effective physical exposure assessment for manual materials handling (MMH) is essential for identifying activities that increase the risk of work-related musculoskeletal disorders and for guiding ergonomic interventions. However, existing methods are labor-intensive and often fail to capture task variability or to effectively estimate task timing characteristics. We evaluated the use of vision-language models (VLMs) to automatically and non-invasively classify eight MMH tasks and specific task conditions (i.e., hand configuration and lifting origin), and to detect task start and end times, using regular RGB video streams. We obtained task classification accuracies of ∼82-85%, accuracies for classifying lifting origin of ∼94-98%, and mean absolute start and end time errors <0.5 s, superior to prior work in some cases. Classification performance for hand configuration, though, was more variable. These findings demonstrate the potential of VLMs as a practical and scalable tool for physical exposure assessment of MMH tasks.
Passive back- and arm-support exoskeletons have effects on physical demands and user perceptions in simulated manual mining tasks that are generally beneficial but are both device- and task-specific
Akinwande, Feyisayo; Kim, Sunwook; Nussbaum, Maury A. (Elsevier, 2026-05)
Back-support (BSE) and arm-support exoskeletons (ASEs) can reduce physical demands during occupational tasks, yet their effectiveness in manual mining tasks remains unknown. We evaluated two BSEs (HeroWear, Laevo Flex) and two ASEs (EVO, Paexo Shoulder) in two lab-based studies (n = 18 each), each including a no-exoskeleton condition. The BSE study involved simulated cable hanging/installation, core box lifting, and overhead wire-mesh installation; the ASE study included these tasks plus overhead drilling. Both BSEs significantly reduced peak trunk extensor activity during lifting (∼15-30%) and overhead installation (∼9-21%), although perceptions differed: HeroWear reduced upper-back exertion, whereas Laevo Flex increased waist/hip discomfort. Paexo Shoulder significantly reduced total shoulder-muscle activity during cable hanging (∼23%), core box lifting (∼22%), and wire-mesh installation (∼24%), while EVO yielded significant reductions during wire-mesh installation (∼17%). Both ASEs reduced perceived exertion across most body regions with minimal discomfort. Importantly, both BSE and ASE benefits were generally consistent across task-specific conditions, suggesting that their effectiveness is robust across varying task demands. Despite their potential for musculoskeletal disorder prevention, field evaluation is recommended.
A new epidemics–logistics model: Insights into controlling the Ebola virus disease in West Africa
Büyüktahtakın, İ. Esra; des-Bordes, Emmanuel; Kıbış, Eyyüb Y. (Elsevier, 2018-03)
Compartmental models have been a phenomenon of studying epidemics. However, existing compartmental models do not explicitly consider the spatial spread of an epidemic and logistics issues simultaneously. In this study, we address this limitation by introducing a new epidemics–logistics mixed-integer programming (MIP) model that determines the optimal amount, timing and location of resources that are allocated for controlling an infectious disease outbreak while accounting for its spatial spread dynamics. The objective of this proposed model is to minimize the total number of infections and fatalities under a limited budget over a multi-period planning horizon. The present study is the first spatially explicit optimization approach that considers geographically varying rates for disease transmission, migration of infected individuals over different regions, and varying treatment rates due to the limited capacity of treatment centers. We illustrate the performance of the MIP model using the case of the 2014–2015 Ebola outbreak in Guinea, Liberia, and Sierra Leone. Our results provide explicit information on intervention timing and intensity for each specific region of these most affected countries. Our model predictions closely fit the real outbreak data and suggest that large upfront investments in treatment and isolation result in the most efficient use of resources to minimize infections. The proposed modeling framework can be adopted to study other infectious diseases and provide tangible policy recommendations for controlling an infectious disease outbreak over large spatial and temporal scales.
Ebola Control Needs Integrated Epidemics–Logistics Optimization
Büyüktahtakın, İ. Esra (2026-05-22)
The rapidly expanding Bundibugyo Ebola outbreak in the Democratic Republic of the Congo and Uganda underscores a persistent implementation gap: risk signals do not automatically become timely, feasible action. This Comment argues that integrated epidemics–logistics optimization can help decision makers allocate treatment, isolation, diagnostics, transport, and reserve capacity under uncertainty, scarce resources, and socio-political constraints. The analysis highlights four operational priorities: early targeted deployment, dynamic reallocation, equity-aware response, and protected surge capacity.
An adaptive K-means and reinforcement learning (RL) algorithm to effective vaccine distribution
Cibaku, Elson; Büyüktahtakın, İ. Esra (Elsevier, 2026-01)
We present a new adaptive reinforcement learning (RL) approach, integrated with a K-means clustering algorithm and guided by simulated annealing, to address the capacitated vehicle routing for vaccine distribution (CVRVD) problem. This integrated method provides an efficient and scalable solution for optimizing vaccine distribution logistics. By incorporating cost factors related to travel distance, inventory levels, and penalty terms – while adhering to delivery time windows – our approach improves both operational efficiency and vaccine allocation effectiveness. Experimental results demonstrate that our K-means supported RL algorithm significantly outperforms traditional solvers in tackling this NP-hard problem, particularly in large-scale scenarios. Specifically, our approach can efficiently solve CVRVD instances with up to 1,000 facilities—scenarios that are computationally intractable for exact methods. We demonstrate the effectiveness of the adaptive K-means supported RL algorithm using data from New Jersey, USA, where facility-level vaccination data were available through the state's Immunization Information System. Beyond vaccine distribution, our method has broad applicability in logistics and transportation, enabling more efficient and cost-effective allocation of critical resources such as vaccines and medical supplies.
Discovering heuristics with Large Language Models (LLMs) for mixed-integer programs: Single-machine scheduling
Cetinkaya, Ibrahim Oguz; Büyüktahtakın, İ. Esra; Shojaee, Parshin; Reddy, Chandan K. (Pergamon-Elsevier, 2026-02)
Our study contributes to the scheduling and combinatorial optimization literature with new heuristics discovered by leveraging the power of Large Language Models (LLMs). We focus on the single-machine total tardiness (SMTT) problem, which aims to minimize total tardiness by sequencing n jobs on a single processor without preemption, given processing times and due dates. We develop and benchmark two novel LLM-discovered heuristics, the EDD Challenger (EDDC) and MDD Challenger (MDDC), inspired by the well-known Earliest Due Date (EDD) and Modified Due Date (MDD) rules. In contrast to prior studies that employed simpler rule-based heuristics, we evaluate our LLM-discovered algorithms using rigorous criteria, including optimality gaps and solution time derived from a mixed-integer programming (MIP) formulation of SMTT. We compare their performance against state-of-the-art heuristics and exact methods across various job sizes (20, 100, 200, and 500 jobs). For instances with more than 100 jobs, exact methods such as MIP and dynamic programming become computationally intractable. Up to 500 jobs, EDDC improves upon the classic EDD rule and another widely used algorithm in the literature. MDDC consistently outperforms traditional heuristics and remains competitive with exact approaches, particularly on larger and more complex instances. This study shows that human-LLM collaboration can produce scalable, high-performing heuristics for NP-hard constrained combinatorial optimization, even under limited resources when effectively configured.
An Interpretable Ensemble Heuristic for Principal-Agent Games with Machine Learning
Baswapuram, Avinashh Kumar; Chen, Chen; Cai, Wenbo; Büyüktahtakın, İ. Esra (2026-04)
This paper addresses the challenge of enhancing public policy decision-making by efficiently solving principal-agent models (PAMs) for public-private partnerships, a critical yet computationally demanding problem. We develop a fast, interpretable, and generalizable approach to support policy decisions under these settings. We propose an interpretable ensemble heuristic (EH) that integrates Machine Learning (ML), Operations Research (OR), and Game Theory. First, we reformulate a PAM as a mixed-integer program to improve efficiency. Next, we solve thousands of PAM instances under varying config- urations to generate training data for ensemble tree-based ML models that identify key solution patterns. These patterns form a hierarchical heuristic that provides feasible and interpretable solutions. We demonstrate the EH’s efficacy in managing the Emerald Ash Borer (EAB) infestation, an urgent public-policy threat to U.S. ash trees. Empirical results show that the EH produces high-quality solutions with 1-2% optimality gaps while significantly reducing computational time compared to exact optimization. Furthermore, the heuristic explains predictions using an average of 4.5 of 9 input features, enhancing transparency. Our findings demonstrate that the EH promotes rapid, informed, and accountable policy decisions by balancing interpretability with computational efficiency. Practically, it supports real-time simulations for stakeholders without specialized ML or OR expertise. Methodologically, it demonstrates a robust integration of optimization and machine learning to solve complex policy models. Beyond the EAB application, this approach provides a scalable framework for real-time decision support where transparency and justification are paramount.
Deep Learning for Sequential Decision Making under Uncertainty: Foundations, Frameworks, and Frontiers
Büyüktahtakın, İ. Esra (2026-04)
Artificial intelligence (AI) is moving increasingly beyond prediction to support decisions in complex, uncertain, and dynamic environments. This shift creates a natural intersection with operations research and management sciences (OR/MS), which have long offered conceptual and methodological foundations for sequential decision-making under uncertainty. At the same time, recent advances in deep learning, including feedforward neural networks, LSTMs, transformers, and deep reinforcement learning, have expanded the scope of data-driven modeling and opened new possibilities for large-scale decision systems. This tutorial presents an OR/MS-centered perspective on deep learning for sequential decision-making under uncertainty. Its central premise is that deep learning is valuable not as a replacement for optimization, but as a complement to it. Deep learning brings adaptability and scalable approximation, whereas OR/MS provides the structural rigor needed to represent constraints, recourse, and uncertainty. The tutorial reviews key decision-making foundations, connects them to the major neural architectures in modern AI, and discusses leading approaches to integrating learning and optimization. It also highlights emerging impact in domains such as supply chains, healthcare and epidemic response, agriculture, energy, and autonomous operations. More broadly, it frames these developments as part of a wider transition from predictive AI toward decision-capable AI and highlights the role of OR/MS in shaping the next generation of integrated learning–optimization systems.
Adaptation to a Whole-Body Powered Exoskeleton: Human-Exoskeleton Coordination During Load-Handling Tasks
Park, Hanjun; Kim, Sunwook; Nussbaum, Maury A.; Srinivasan, Divya (Springer, 2026-03)
Whole-body powered exoskeletons can augment human performance and reduce physical strain in occupational settings, but little is known about how users adapt to these complex devices during practical work scenarios. We compared novice and experienced users during simulated, occupationally relevant load-handling tasks. Six novice users completed exoskeleton familiarization and stationary load-handling tasks in three sessions while five experienced users performed the tasks once. Task performance, biomechanical demands, and perceived workload were compared in each novice session vs. the experienced group. Novice performance improved substantially across sessions, with task completion time reduced by nearly 50% and movement jerk by 30%. However, performance gaps still persisted in session three, compared to the experienced users. Novices also used consistently lower angular velocities (up to 52% lower) and adopted greater hip flexion throughout the sessions. In contrast, differences in shoulder flexion, muscle activity, perceived exertion, and workload diminished more rapidly, with novices approaching experienced levels by session three. Novice users adapted to using a powered exoskeleton over multiple sessions, especially in movement patterns and muscle activation, but differences in task completion time, jerk index, and angular velocities indicated that novices did not attain the skilled coordination and efficiency of experienced users after three sessions. Our results highlight the likely need for extended familiarization and training for the current powered exoskeleton design and provide baseline data for the novice learning curve in occupational settings.
Not All Noises Are Equal: Investigating Auditory Distraction in Emergency Care Using the Tesseract Simulation Platform
Du, David; Lau, Nathan; Ojeifo, Olumide A.; Upthegrove, Tanner; Baber, Adam; Jones, Nathan A.; Parker, Sarah H. (SAGE Publications, 2025-09)
Auditory distractions in clinical environments can impair performance, yet their impact in emergency department (ED) waiting rooms remains understudied. This study investigates how distinct noise types (baby crying, conversations, and equipment alarms) and temporal patterns (continuous vs. intermittent) influence nursing triage performance. Thirty-two ED nurses completed standardized triage tasks within the Tesseract, an immersive audio-visual simulation platform replicating ED waiting room conditions. Preliminary results from 16 participants suggest that the effects of noise depend on both acoustic features and temporal structures. Subjective perceptions of distraction did not consistently align with measured outcomes. This work provides early evidence that auditory distractions influence clinical task execution in complex, task- and context-specific ways, underscoring the need for targeted mitigation strategies and soundscape-aware simulation training.
Sustainable Timber Supply Chain Optimization
Adams, Irma; Canuel, Austin; Lee, Minseo; Büyüktahtakın, İ. Esra (2026-02-23)
The U.S. timber supply chain faces mounting challenges related to capacity constraints, sustainability, and supply resilience at a time when federal policy calls for a rapid expansion of domestic timber production. Following the March 2025 executive order to reduce reliance on foreign timber imports, achieving near-term production targets requires a nationwide redesign of supply chain infrastructure under significant data and operational uncertainty. This study develops a data-driven optimization framework to support short-term, actionable planning for the U.S. timber supply chain. We propose a hybrid machine learning–mixed-integer linear programming (ML–MILP) model that captures the flow of timber from mills through distribution centers to demand points, with the objective of minimizing total transportation and facility-opening costs. U.S.–wide implementation is complicated by incomplete and fragmented data, particularly for mill counts, production levels, and facility locations. To address these gaps, we leverage machine learning models, including gradient boosting, ridge regression, and weighted K-Means clustering, to reconstruct a comprehensive national dataset and generate candidate distribution center locations informed by socioeconomic and environmental factors. The resulting MILP generates an infrastructure and flow plan and is evaluated through sensitivity and scenario-based analyses reflecting demand growth, transportation disruptions, and disaster impacts. Results highlight the dominant role of transportation costs, diminishing returns to capacity expansion, and heightened vulnerability in the South and West regions. Overall, the proposed framework provides policymakers and industry stakeholders with a scalable, sustainability-oriented decision-support tool for guiding domestic timber supply chain expansion under evolving policy objectives.
Cross-Attention Guided Data Sharing for Knowledge Transfer in Robotic AI Systems
Liu, Hui; Zeng, Yingyan; Qiao, Helen; Piliptchak, Pavel; Jin, Ran (2026)
Cross-robot transfer learning is crucial for building robotic AI systems that can generalize across diverse platforms and tasks by utilizing heterogeneous datasets. However, not all source samples are effective to improve the accuracy of the target AI task; while incompatible samples may lead to negative transfer and degrade model performance. This challenge is particularly in a connected robot fleet where robots differ in configurations, sensors, but are connected via Industrial Internet for sequential or parallel tasks. To address this, we propose Cross-Attention guided Proximal Policy Optimization (CAPPO), a reinforcement learning-based sample selection framework that adaptively identifies the most valuable source samples for a given target AI modeling task. Our method employs cross-attention mechanisms to capture fine-grained relevance between source and target samples, constructing informative state representations for a PPO-based selection policy. A task-driven reward function based on downstream performance improvement is created to enable the agent to learn efficient and adaptive selection strategies. Experimental results on a connected robotic fleet with different AI tasks show that our method consistently outperforms existing baselines under low-budget settings, demonstrating strong and robust knowledge transfer performance to train new robotic AI models.

Browse

Recent Submissions