Adaptive Virtual Labs for Engineering Skills under Resource Constraints

Thesis Abstract

The rapid expansion of engineering education demands high-fidelity practical learning experiences, yet access to traditional laboratory facilities is frequently constrained by budget, equipment scarcity, and geographic limitations, particularly in regional or underfunded institutions. This study addresses the problem of delivering equitable, hands-on engineering competencies under resource constraints by developing and evaluating adaptive virtual laboratories that tailor simulation fidelity and hardware-in-the-loop components to learner context, equipment availability, and time. The aim is to investigate whether adaptive virtual labs can (i) preserve core experimental learning outcomes, (ii) reduce inequitable access to practical training, and (iii) improve student engagement and self-regulated learning in resource-limited environments. The specific objectives are (a) to design an adaptable virtual lab platform for core engineering domains (mechanical, electrical, and control systems) leveraging cloud-based simulation, remote instrumentation, and modular hardware proxies; (b) to establish an adaptive scaffolding algorithm that modulates simulation fidelity, data resolution, and experimental complexity based on student proficiency, access constraints, and learning goals; (c) to evaluate the platform’s impact on learning gains using a quasi-experimental design across three faculties with diverse resource profiles; (d) to examine user experience, perceived usefulness, and engagement through mixed-methods data; and (e) to formulate a model of resource-constrained-practice equivalence to traditional laboratories. The methodology adopts a mixed-methods, multi-site design over two academic cycles. The population comprises undergraduate and postgraduate engineering students enrolled in electrical, mechanical, and mechatronics programs at three public universities. A total sample of 480 students will be selected via stratified random sampling across faculties, with 160 participants per university and balanced representation from year levels. The intervention group (n ? 240) will use the adaptive virtual labs for selected experiments over a 12-week semester, while the control group (n ? 240) will rely on conventional labs and demonstrations. Data collection instruments include (i) standardized pre- and post-tests assessing practical competencies in core experiments (e.g., circuit assembly, PID control tuning, and fluid dynamics visualization); (ii) the User Experience Questionnaire (UEQ) and System Usability Scale (SUS) to capture usability and engagement; (iii) log analytics capturing time-on-task, fidelity levels, and resource utilization; (iv) semi-structured interviews with a purposive subsample (n = 40) to extract in-depth perceptions of adaptability, reliability, and perceived impact on learning; and (v) focus groups with instructors (n = 12) to glean implementation feasibility. Validity and reliability will be established through pilot testing, Cronbach’s alpha checks (target >0.80 for scales), and test–retest procedures. Data analysis will proceed as follows descriptive statistics to profile participants; inferential statistics using ANCOVA to compare learning gains while controlling pre-test scores; regression analysis to explore predictors of achievement and engagement; thematic analysis of interview data following Braun and Clarke’s approach to identify patterns related to adaptability, accessibility, and pedagogical impact; and a convergent mixed-methods integration to triangulate quantitative and qualitative findings. A deterministic analytical framework will be complemented by a model-based evaluation of the adaptive scaffolding algorithm, employing a Markov decision process to optimize sequence and fidelity adjustments under varying resource constraints. Expected findings include significant improvements in practical competencies for the intervention group, with effect sizes in the range of partial eta-squared 0.06–0.12, and higher engagement scores compared with controls. The adaptive fidelity mechanism is anticipated to preserve core learning outcomes despite reduced hardware availability, demonstrating that tailored simulators and hardware proxies can substitute for full-scale laboratories without compromising skill transfer. Furthermore, the study is expected to reveal nuanced differences by domain (electrical vs mechanical vs mechatronics) and by student background, informing targeted design recommendations. The contribution to knowledge lies in (i) advancing a validated framework for resource-aware virtual laboratory design in engineering education, (ii) providing empirical evidence on the efficacy of adaptive fidelity and hardware proxies in achieving equivalence or superiority to traditional labs under constraints, and (iii) offering a practical blueprint for scalable deployment across resource-diverse institutions. The main conclusion will articulate that adaptive virtual labs, when underpinned by a dynamic instructional scaffold and grounded in learner analytics, can effectively mitigate resource constraints while sustaining or enhancing practical engineering competencies. Recommendations include scaling the platform to additional domains, integrating live remote instrumentation where feasible, and developing policy guidelines for resource allocation and faculty development to foster broader adoption.