TThe past decade has marked a decisive transformation in educational environ-ments, driven by rapid advancements in digital technologies and accelerated by the COVID-19 pandemic. Since 2016, broadband infrastructure, learning man-agement systems, and immersive instructional tools such as virtual and augment-ed reality have redefined teaching and learning processes worldwide. The sudden shift to remote instruction in 2020 further intensified reliance on digital platforms, fostering hybrid synchronous–asynchronous learning modes in the post-pandemic period. Within this context, engineering education—and particularly Computer-Aided Design (CAD) instruction—has undergone significant peda-gogical and technological evolution.

This study examines the development of CAD education over the last decade by conducting a bibliometric analysis of literature indexed in the Scopus database be-tween 2016 and 2026. Using a systematic keyword search and RIS-file extrac-tion, the dataset was processed in VOSviewer to generate keyword co-occurrence networks and overlay visualizations. The analysis identified 828 relevant publica-tions and revealed twelve thematic clusters reflecting pedagogical foundations, digital design tools, emerging technologies, artificial intelligence integration, and professional training. Dominant research themes include engineering education, design instruction, 3D modeling, and Industry 4.0-aligned digital competencies. Emerging topics, such as open-source platforms, benchmarking, machine learn-ing, and data-driven design, indicate increasing convergence between CAD edu-cation and advanced computational methods.

The findings highlight a research transition from software-centric approaches to-ward holistic, technology-enhanced pedagogies that promote digital literacy, ex-periential learning, and interdisciplinary design. However, gaps remain in linking CAD competencies to broader educational transformations such as lifelong digital skills, competency-based learning, and sustainability-driven design practices.

This paper presents a modeling environment for global human-centric digi-tal strategies that emphasize technologically enhanced production. Building on Industry 4.0 foundations, it explores how AI, collaborative robots, digi-tal twins, and sensor-driven simulations can be leveraged to optimize ergo-nomics, productivity, and economic efficiency in work processes. The ap-proach integrates automated analysis tools, intuitive drag-and-drop inter-faces, and a comprehensive robot library to enable skill-based human-robot task allocation and realistic, safe simulations. Designed for education, the system fosters exploratory, problem-based learning, aligning cognitive psy-chology principles with hands-on experiences to enhance understanding of complex cyber-physical systems. By combining technological, ergonomic, and didactic perspectives, the method supports the acquisition of Future Skills, promotes active learner engagement, and prepares students for the challenges of the human centered industrial paradigms.

CONTEXT

Modern military field operations face critical bandwidth constraints due to reliance on satellite uplinks, tactical radio systems, and the need for low electromagnetic signatures. Despite these limitations, legacy network traffic including advertisements, telemetry, background updates, and tracking scripts consume substantial portions of available bandwidth, negatively impacting mission-critical communications. While existing research has explored Software-Defined Networking (SDN) approaches for agile routing and load balancing in constrained military environments, a significant gap exists in solutions targeting traffic filtering at the DNS level to reduce overhead and enhance security posture. This study addresses this gap by examining DNS-based traffic optimization as a complementary approach to existing military networking strategies, building upon proven civilian applications of DNS filtering technologies that have demonstrated bandwidth savings of 10-30% in non-military contexts.

PURPOSE OR GOAL

This research investigates whether deploying Pi-hole—an open-source DNS-level filtering solution—on low-power Raspberry Pi hardware within military field networks can effectively minimize non-essential traffic, reduce latency, and augment security in bandwidth-constrained operational environments. The study seeks to answer the primary research question: Can DNS-based filtering provide measurable improvements in bandwidth utilization, network performance, and security posture when integrated with enterprise-grade routers in simulated military field conditions? The motivation stems from the critical need to maximize available bandwidth for mission-critical communications while simultaneously reducing security vulnerabilities associated with advertising networks, tracking services, and malicious domains. This work aims to demonstrate a scalable, cost-effective solution that complements existing SDN capabilities and can be rapidly deployed in tactical environments where every transmitted byte impacts operational effectiveness.

APPROACH

The study employed a controlled deployment methodology using Raspberry Pi Model 3B hardware (selected for its low power draw of approximately 5W and rugged potential) integrated with an ASUS RT-AX1800U enterprise-grade router in a simulated field environment. The Pi-hole DNS filtering solution was configured as the primary DNS resolver for all client devices, with security enhancements including DNSSEC validation, query rate limiting, and curated military-specific blocklists. The network topology utilized a hub-and-spoke architecture with Cat-6 wired connections, DHCP configuration assigning Pi-hole as primary DNS, and QoS rules prioritizing mission-critical traffic. Data collection focused on four primary metrics: total bandwidth utilization, external DNS query frequency, web page load times, and simultaneous connection counts. Performance was measured by comparing network behavior with Pi-hole active versus inactive states across multiple simulated mission scenarios. Security effectiveness was evaluated by testing detection and blocking capabilities against known malicious domains from military threat listings. The deployment incorporated network segmentation via VLANs to isolate mission traffic, and comprehensive logging enabled detailed analysis of DNS query patterns and potential security incidents.

Integrating advanced technologies into educational environments is essential for training students in industrial automation. In this context, the Polytechnic School of Engineering of Ferrol (EPEF) has incorporated Phoenix Contact PLCnext technology, thanks to its membership in the EduNet - International Education Network. PLCnext technology represents a significant evolution from traditional controllers, expanding their capabilities beyond the languages defined by the IEC 61131 standard. This versatility allows for the integration of modern languages, facilitating the development of more advanced applications adapted to the current challenges of industrial automation. In the educational context, PLCnext offers a hybrid environment that promotes practical learning, experimentation, and the acquisition of cross-cutting skills in programming, control, and data management. The main objective of this work was to design and implement a comprehensive data acquisition and control system for laboratory level plants, using PLCnext controllers as the system's decentralized cores. The aim was to demonstrate the viability of this technology in educational environments by enabling remote control, SCADA visualization, and data logging. In addition, the flexibility of PLCnext to integrate tools such as MATLAB/Simulink in a laboratory environment was demonstrated.

Abstract. This article presents a new framework designed to improve image interpretation in four high school science disciplines through the use of open-weight language models optimized for resource-constrained environments. It also makes a significant contribution to the field of educational AI. While most previous studies focus on proprietary or hybrid models, evaluated using benchmarks such as VisioMath, ScienceQA, SceMQA, or MMMU, our study proposes an alternative

approach. We optimized four open-weight models—Mistral-7B, LLaVA-1.5-7B, Kosmos-2, and Qwen2-VL—selected for their lightweight architecture, low memory footprint, and robustness. The optimization focused on low-rank adaptation (LoRA) and 4-bit quantization in a resourceconstrained environment, specifically to improve the understanding of scientific images. The results of Bench Low, distinguishing between multiple-choice (MC) and free-response (FR) questions, reveal significant variations depending on the type of task, model, and discipline. LLaVA-1.5-7B stood

out in particular for its superior performance, achieving near-perfect accuracy on multiple-choice questions (96.9%) and the highest score on open-ended questions (69.1%), with an overall average of 79.3%. In contrast, Kosmos-2 achieved the lowest results (40.8% overall), with high variability between disciplines, reflecting uneven coverage of training data. For reproducibility purposes, the source codes are fully available on Github: https://github.com/mouazmikail/Bench-Lowv1/tree/master