VirtLab: Digital Laboratory Systems
- VirtLab is a class of digital laboratory systems that replicate and enhance traditional experiments through software-driven environments.
- It integrates immersive VR, haptics, and digital twins to facilitate remote training, simulation, and real-time data analysis.
- Architectures in VirtLab streamline experiment management, metadata capture, and collaborative workflows, advancing innovative scientific research.
VirtLab denotes a class of virtual laboratory systems in which experimentation, observation, analysis, and collaboration are carried out in software while preserving core laboratory functions. In the literature, the term encompasses immersive educational laboratories with haptics or virtual reality, digital-twin environments for remote practical training, browser-based reconstructions of real facilities, virtual and remote control interfaces for physical apparatus, computational experiment management systems, and domain-agnostic workflow environments for scientific discovery (Hamza-Lup et al., 2019, Palmer et al., 2021, Adamidi et al., 1 Apr 2025, Sevilla-Salcedo et al., 8 Jul 2025). Across these uses, the unifying feature is not a single interface technology, but the transfer of laboratory practice into a controllable digital environment that can support experimentation, iteration, and, when needed, linkage to physical instruments or shared computational infrastructure.
1. Conceptual scope and historical development
One of the earliest formulations close to the modern VirtLab idea appears in computational astrophysics, where virtual laboratories were described as environments in which scientists conduct experiments on simulated systems that cannot be reproduced physically, such as stars. In that account, the historical limitation was that researchers remained “behind the keyboard,” whereas 3D online environments such as Second Life and Qwaq Forums made it possible to enter the virtual lab as avatars, inspect simulations spatially, and collaborate in persistent rooms with shared displays, blackboards, and documents (0712.1655).
Subsequent work broadened the meaning of virtual laboratory beyond immersive 3D presence. Empirica defined a virtual lab as infrastructure for high-throughput, macro-level online experiments, emphasizing parameterizable designs, reusable protocols, and rapid development rather than spatial immersion (Almaatouq et al., 2020). SCHEMA lab extended the idea to computational reproducibility by treating the virtual laboratory as a workspace for grouping multiple executions into experiments, capturing rich metadata, and managing tasks and workflows over their life cycle (Adamidi et al., 1 Apr 2025). VAILabs made the broadest definition explicit: “virtual” was stated not to mean virtual reality, but rather a digital counterpart to a physical laboratory in which research activities are orchestrated through modular workflows and interfaces to digital or physical components (Sevilla-Salcedo et al., 8 Jul 2025).
Taken together, these works suggest that “VirtLab” functions as an umbrella label for several related but non-identical laboratory paradigms.
| Formulation | Representative papers | Primary emphasis |
|---|---|---|
| Educational and visuo-haptic labs | (Hamza-Lup et al., 2019, Haldolaarachchige et al., 2020, Haldolaarachchige et al., 2020) | Instructional experiments, simulation, analysis |
| Immersive VR and digital twins | (Palmer et al., 2021, Perez et al., 17 Mar 2025, Fan et al., 13 Mar 2026) | Embodied practice, facility familiarization, remote training |
| Computational research workbenches | (Almaatouq et al., 2020, Adamidi et al., 1 Apr 2025, Sevilla-Salcedo et al., 8 Jul 2025) | Experiment orchestration, metadata, reproducibility |
| LLM-based team simulation labs | (Almutairi et al., 6 Aug 2025, Almutairi et al., 9 Oct 2025) | Study of coordination in spatial, time-evolving environments |
2. Pedagogical rationale and laboratory pedagogy
In educational usage, VirtLab is typically justified as a response to the loss of experimentation in remote teaching. The virtual Physics I and Physics II lab manuals converted conventional undergraduate laboratories into simulation-based and video-analysis-based courses using only open educational resources, Microsoft Excel for analytical and graphical computation, a Learning Management System for organization and recording, and live video conferencing for synchronous delivery. Both manuals were explicitly designed to simulate in-person physical laboratory experiments, to support synchronous and asynchronous teaching modes, and to assess higher-order outcomes such as understand, apply, analyze, and evaluate through detailed lab reports and end-of-semester written examinations. The Physics II manual states that the virtual format was confirmed to be as effective as the in-person physical lab class (Haldolaarachchige et al., 2020, Haldolaarachchige et al., 2020).
Other educational VirtLab formulations place greater emphasis on preparation, multimodality, and procedural rehearsal. The haptic physics paper argues that authentic learning requires experimentation and action, and presents touch-enabled simulations as especially valuable for concepts that are difficult to convey visually alone, including static versus kinetic friction, the Coriolis effect, and gyroscopic precession (Hamza-Lup et al., 2019). In biology, OnLabs was used as a 3D, game-like microscopy preparation tool for 43 third-year undergraduate distance-learning students at the Hellenic Open University. Students in the enriched condition received a 1-hour Skype session, observed a full microscopy procedure in OnLabs, then repeated the procedure themselves before attending the physical wet lab. The study concluded that the virtual lab improved baseline preparation and was a valuable supplement rather than a replacement for conventional instruction (Paxinou et al., 2017).
The systematic review literature frames these educational systems as part of a broader shift toward interactive virtual reality laboratories. One review of work from 2016 to 2021 identified 32 relevant papers for year-based analysis and found that most studies evaluated learning with questionnaires, polls, and surveys rather than problem-solving tasks. It also proposed a six-component architecture—Connection Plugin, Data Analysis Module, Mathematical Models, System Database, VR System, and User Module—indicating that virtual laboratories are typically conceived as layered systems that combine interaction, data handling, and scientific modeling rather than graphics alone (Rahman et al., 2022).
3. Interaction modalities and embodied learning
A central line of VirtLab research concerns the sensory and motor form of laboratory interaction. In the haptic physics work, the simulations were implemented with H3D API, X3D, and Python, and used the Novint Falcon as the principal force-feedback device. The friction simulator restricted motion to one dimension along an incline while preserving a 3D visual environment, arrows for force direction and magnitude, and a heads-up display of numerical values. The normalized gain metric was reported as
with a gain of 0.182 for the haptic lab group and -0.011 for the traditional lab group in a study of 86 participants. In the Coriolis study, 24 undergraduate physics students were divided into four groups; the visuo-haptic groups showed about a 15% advantage in quiz scores over the reading-and-video-only groups, while the visual-only simulation improved scores by about 10% (Hamza-Lup et al., 2019).
Virtual reality laboratories have also been used to change the epistemic structure of laboratory work. Novel Observations in Mixed Reality (NOMR) labs placed students in a fictitious universe with consistent but unknown physics, thereby eliminating the possibility of merely confirming a known result. In Fall 2022, these labs were studied in a 100-level calculus-based electromagnetism course and a 200-level introductory experimental physics course at the University of Washington. The reported outcome was that students in both populations became more expertlike in their epistemology about experimental physics and stronger in self-efficacy; all five self-efficacy items shifted positively at a 99.7% confidence level or better, with . Through the lens of flow, the authors argued that the productive engagement persisted after the novelty of VR wore off, which they estimated at about two weeks (Canright et al., 2023).
Embodied engineering variants of VirtLab make bodily congruence a design principle. A 2025 framework for virtual laboratory environments in mechanical and materials engineering used Unity 3D, C#, the XR Interaction Toolkit, and OpenXR, and organized the system as an event-driven directed graph built around an Observer Pattern. Students physically grabbed specimens, positioned them in a tensile tester, observed a generated stress-strain graph, or used “handle diversion” to apply tension and compression in a Poisson’s ratio module. After a pre-test, a self-paced VLE experience of about 10 minutes per student, and a post-test, score increases in the embodied condition were reported as 49% for Question 1, 133% for Question 2, and 65% for Question 3; the cited previous non-embodied VR version had shown a 3% decrease, a 78% increase, and a 47% increase on the same questions (Perez et al., 17 Mar 2025).
Surveying education provides a further example of technically explicit embodied interaction. VRISE, implemented on Meta Quest 3 with Unity 2022 LTS and OpenXR support, used single exponential smoothing to stabilize controller motion:
with . Across five differential-leveling attempts, elevation error improved from 0.4% to 0.05%, completion time decreased from about 320 s to about 265 s, and interaction count fell from 30 actions to 15 actions. The platform also included aerial surveying through waypoint-based drone navigation, suggesting that VirtLab can be used to model fine motor control, spatial reasoning, and task efficiency within the same immersive environment (Udekwe et al., 30 Jul 2025).
4. Digital twins, WebVR, and remote apparatus
A major architectural strand in VirtLab research uses digital twins to reduce authoring cost and to preserve procedural structure across devices. The digital-twinning remote-laboratory work proposed a Digital Twin Builder for lecturer authoring, three processing pipelines—data, geometry, and process—and a Digital Twin Player for delivery. In the electrical laboratory case study, the tutorial objective was to measure the armature resistance of a DC motor by connecting a circuit, switching on the DC supply, increasing the input voltage slowly from 5V to 40V in 5V steps, and recording measurements in a table. The authors argued that conventional creation of complex learning scenarios can take over a year, whereas their generic-model approach reduced development time by a factor of 12, from weeks to hours. In the geometry pipeline, training data were generated from 37 mechanical objects, 108 meshes per object, and 19,980 training meshes; the reported low-poly acceptance rate was 97%, and one example model with 106 parts was reduced from roughly 8 hours of manual modeling to 23 minutes on a dedicated server (Palmer et al., 2021).
Web-based digital twins extend this approach from apparatus to full-scale facilities. A 2026 civil-engineering VirtLab reconstructed a laboratory of over 4,000 m² from massive colorized point clouds using a hybrid Unity-Potree WebVR framework. Data were captured with a Leica RTC360 scanner from 22 scan stations, fused in Leica Cyclone, and reduced from about 17.2 GB to about 5.8 GB using distance-based downsampling with minimum point spacing of 5 mm. Potree handled octree-based level-of-detail rendering; Unity handled interaction logic, navigation, hotspots, and emergency evacuation. The resulting system supported first-person exploration, waypoint teleportation, automatic tours, safety and equipment information, and evacuation rehearsal. In 35 valid questionnaires, 100% of participants rated usability as good or excellent, and 74.3% found navigation and spatial presentation extremely helpful (Fan et al., 13 Mar 2026).
VirtLab can also couple simulation and real hardware through a unified interface. A virtual and remote robotic laboratory based on Easy Java Simulations, MATLAB, and LabVIEW allowed students to use the same graphical interface either to interact with a simulated robot or to tele-operate a real LEGO Mindstorms NXT robot over the Internet. The environment was represented by a 300 × 400 MATLAB matrix called world, updated every 200 ms through ComputeWorld(world, x, y, th, d), while automatic control used ComputeControl(world, x, y, th, vx, vy, w, d, t). A camera mounted about 3 m above a 3 × 4 m² workspace operated at around 17 fps, and motor speed was limited to 30% of maximum for safety. Students were required to succeed in simulation before entering the real mode, and 74% reported feeling more comfortable using the virtual lab first (Chaos et al., 2024).
5. Computational and domain-agnostic research virtual laboratories
In computational research, VirtLab often denotes infrastructure for designing, orchestrating, and preserving experiments rather than immersive 3D scenes. Empirica exemplifies this usage. It was introduced as a modular virtual lab for online, synchronous, interactive human-participant experiments, intended to resolve the trade-off between usability and functionality through a “flexible defaults” strategy. Its architecture centers on server-side callbacks for experiment logic, client-side interfaces for participant interaction, and a GUI admin interface for experiment configuration and management. Experiments are structured hierarchically as players in a game, games containing rounds, and rounds containing stages, while full protocols can be specified in YAML and tested with a bot API (Almaatouq et al., 2020).
SCHEMA lab extends the computational notion of VirtLab toward reproducibility. It comprises a SCHEMA lab front-end and a SCHEMA api back-end, with TESK mediating task execution on a Kubernetes cluster. The system lets researchers group task and workflow executions into experiments, understood as computational efforts, and manage them through create, list, retrieve, update, delete, and task-assignment operations. It captures execution configuration, environment variables, input and output mount points, volumes, workflow order, timestamps, unique UUIDs, performance metrics, resource consumption, execution footprint, and quota usage. This formulation makes the virtual laboratory a metadata-rich management layer above containerized execution rather than merely a runner of tasks (Adamidi et al., 1 Apr 2025).
VAILabs makes the research-workflow interpretation fully explicit. It defines Virtual Laboratories as domain-agnostic digital environments for scientific discovery, not as virtual reality systems. The architecture is organized around modules, plugins, and cores, with an XML-based workflow description. Essential modules include Initialiser, Output, Data Processing, Decision Making, Environment, Modeling, and User Interaction, and workflows may include loops, conditional repetition, and mid-run human intervention. The proof-of-concept mappings include perovskite solar-cell optimization over the space
where only about 1.8% of the discretized space was sampled; robot co-adaptation with an inner reinforcement-learning loop and an outer Bayesian-optimization loop; and de novo drug design with REINVENT, active learning, and binary 0/1 human feedback. This suggests that VirtLab can describe discovery pipelines as much as rendered spaces (Sevilla-Salcedo et al., 8 Jul 2025).
6. Team simulation, evaluation practice, and recurrent limitations
A recent extension of VirtLab treats the laboratory itself as a configurable environment for studying team behavior with LLM-based agents. One system presents a simulation engine and web interface that allow technical and non-technical users to formulate, run, and analyze team simulations in spatial and temporal settings (Almutairi et al., 6 Aug 2025). A closely related formulation, VirT-Lab, represents environments as a 2D matrix , a graph , and a connectivity structure , while modeling discrete events through tuples such as
Agents are configured with roles, traits, memories, and backstories; memories are stored in a FAISS vector database and retrieved through RAG. In the ASIST-based search-and-rescue comparison, 20 simulations were run, with 10 high-trust and 10 low-trust conditions. Human ground-truth scores were consistently higher than simulated scores across six team-functioning dimensions, with the largest gaps in team coordination and emerging leadership. In a user study with 12 participants, overall ratings were moderate to positive, and experts were more skeptical than novices (Almutairi et al., 9 Oct 2025).
Across the broader literature, evaluation practice remains uneven. The systematic review of interactive VR laboratories found that questionnaires accounted for 61.5% of evaluation approaches, polls for 19.2%, surveys for 11.5%, and problem-solving only 7.7%, which the review interpreted as a weakness in performance-based assessment (Rahman et al., 2022). The same review identified recurrent barriers: gear use problems at 31.6%, understanding time at 21.1%, gear cost and nausea at 15.8% each, connection issues at 10.5%, and technical issues at 5.4%.
Another recurrent issue is overstatement of replacement claims. The biology preparation study presented OnLabs as a supplement to face-to-face wet-lab instruction, not a substitute (Paxinou et al., 2017). The haptic physics work argued that touch-based virtual labs augment rather than replace traditional teaching (Hamza-Lup et al., 2019). The point-cloud WebVR system was evaluated as a complement to conventional on-site laboratory instruction (Fan et al., 13 Mar 2026). At the same time, some research-oriented implementations explicitly reject the assumption that VirtLab is synonymous with immersion: VAILabs states that virtual means software-based orchestration rather than virtual reality, and SCHEMA lab demonstrates a virtual laboratory centered on execution, provenance, and experiment-level structure (Sevilla-Salcedo et al., 8 Jul 2025, Adamidi et al., 1 Apr 2025). A plausible implication is that VirtLab is best understood as an operational pattern for laboratory work in digital form, whose interfaces may range from spreadsheets and LMS sessions to haptic devices, browsers, digital twins, and cluster-backed workflow systems.
