Architecture of complex systems in intelligent immersion – ArchiTOOL

The ArchiTOOL project followed an empirical and iterative scientific approach, structured around several complementary experimental protocols, in line with a methodological triangulation strategy. This approach aimed to assess the effectiveness of the immersive modeling environment from multiple perspectives: user satisfaction, visual notation quality, cognitive load, sense of presence, and collaborative performance. The first phase consisted of an online survey with 118 participants, both experts and non-experts, to compare 2D and 3D representations based on criteria from the Physics of Notations (PoN) framework. The representations were evaluated in terms of visual expressiveness, semiotic clarity, complexity management, and semantic transparency. Statistical analysis helped identify user preferences and the perceived cognitive advantages of immersive notations. Based on these results, an immersive MBSA prototype was developed in Unity3D, integrating tailored visual metaphors for each architectural view (operational, functional, structural, behavioral). This prototype was then evaluated in a controlled individual experiment (n=72), where participants performed modeling tasks either in an immersive or non-immersive environment. Measured variables included cognitive load (NASA-TLX), perceived presence (IPQ), satisfaction (SUS), and PoN criteria related to notation quality. The third phase involved a controlled collaborative experiment (n=48), where participants performed the same modeling tasks in a cooperative setting. This study assessed the impact of immersion on group dynamics, perceived communication quality, and collective performance, comparing the immersive interface with a standard 2D setup. Finally, formative tests with MBSE experts (n=6) were conducted to evaluate system usability, adaptability to professional practices, and gather qualitative feedback on interface limitations. These sessions included user tests with direct observation, semi-structured questionnaires, and interviews. All these methods were integrated into a final triangulation process, combining results from quantitative and qualitative studies. This comprehensive strategy strengthened both the internal and external validity of the project’s conclusions, while offering a nuanced understanding of the immersive environment’s effects depending on user profiles and use contexts.

The ArchiTOOL project enabled a comprehensive evaluation of the immersive MBSA interface across five key dimensions: visual notation quality, cognitive load, sense of presence, user satisfaction, and collaborative dynamics. The methodological approach combined surveys, controlled experiments (individual and collaborative), and expert testing, following a triangulation principle. According to the Physics of Notations (PoN) framework, the immersive 3D interface outperformed traditional 2D notations on several criteria: Perceptual discriminability: generally better in 3D, except in the Fire Forest Detect System case. Other validation methods confirmed the advantage of 3D. Semantic transparency: consistently higher scores in 3D, improving symbol meaning association. Complexity management: immersive 3D and iconic 2D performed similarly and outperformed standard 2D diagrams, with one expert as an exception. Visual expressiveness: 3D demonstrated clear superiority due to its visual richness. Graphic economy: results varied by case; two experts found 3D equivalent or inferior. Presence scores (IPQ) were mixed: overall scores fell into the “unacceptable” zone by IPQ standards, yet some experts reported highly positive perceptions. Individual differences (VR experience, cognitive sensitivity) were significant, suggesting the need to revisit IPQ thresholds for HMD based immersive environments. Cognitive load (NASA-TLX) was generally lower in immersive 3D compared to traditional 2D, with the exception of three experts—again highlighting strong interindividual variability. Usability (SUS) and collaborative performance were also favorable: participants reported high motivation, satisfaction, future use intention, and perceived gains in communication, efficiency, and engagement in collaborative 3D mode. In summary, despite some limitations (VR sickness, user profiles, context dependent effectiveness), the findings point to a significant improvement in user experience with the immersive interface. The delivered system proved robust, validated by both novices and experts, and received overall positive feedback. Project outcomes have been presented at several international conferences —CAD and Applications 2021, CIRP Design 2023 and 2024, and PLM 2025— and in the journal CAD and Applications. Two additional papers are currently under review in international journals.

The ArchiTOOL project provided empirical validation for the relevance of an immersive and intelligent environment in Model-Based Systems Architecting (MBSA). It demonstrated that new forms of visual interaction and collaboration can help reduce cognitive load, improve notation accessibility, and offer enhanced graphic expressiveness. While the results are encouraging, the project also highlighted several limitations that pave the way for further research, particularly regarding the robustness of results, inter-individual variability, collaborative dynamics, and technological acceptability. These research directions are now being developed within the broader strategic framework of the LabCom ANR MIMESIS, in collaboration with Skyreal. Five research perspectives are proposed to extend this work: 1. Impact of graphical representations on multidisciplinary systems architecture. This axis aims to better understand how visual characteristics influence comprehension and collaboration across disciplines. Special attention will be given to the impact of 3D representations—immersive or not—on the detection of inconsistencies, the interpretation of complex models, and cross-disciplinary coordination. 2. Interactive visual metaphors for system modeling. We propose the design of new metaphors to visually represent abstract systems engineering concepts. These metaphors will be dynamic, interactive, adaptive, and integrated with conversational agents to support active model exploration. 3. Co-design of product systems and industrial systems. The goal is to connect product models and industrial processes through 3D interfaces that enable seamless navigation between views. This integrated approach supports global optimization and will be tested in real industrial contexts (e.g., aerospace, space, and energy sectors). 4. Synchronous and asynchronous collaboration in virtual environments. We aim to develop dedicated functions for early-stage collaborative design (e.g., co-editing, contextual transitions, and continuity management with AI agents). Multimodal communication (voice, gesture, annotation) will be explored to enhance engagement and interaction fluidity within distributed teams. 5. Methods to rapidly specify and evaluate virtual environments for system design. We will introduce new functions for rapid prototyping of immersive interfaces, combined with evaluation methods that integrate both objective and subjective criteria. The aim is to adapt virtual environments to the specific needs of each project and user profile. These efforts aim to support the emergence of multi-user 3D environments open to human-AI collaboration, in view of deeply transforming systems engineering practices.

Wang, H., et al. Exploring the potential of virtual reality for model-based systems architecting. 2024. (in press).

Medina-Galvis, S.C., et al. Exploring Model-Based Systems Architecting Digital Threads in Virtual Reality. 2024. (in press).

Pinquié R., et al. Human-Centric Co-Design of Model-Based System Architecture. Procedia CIRP. 2023, 119, 146-151.

The project ArchiTOOL aims at inventing, prototyping, and evaluating an immersive and intelligent virtual environment for architecting complex technological systems. Instead of using domain-specific engineering software, the immersive and interactive environment will provide the architect with the modelling capabilities required to define the various views (operational, specification, functional, behavioural, structural, logic, safety, etc.) of a system architecture in a single virtual space before exporting each viewpoint in a standardised format that will enable domain-experts to continue with a detailed design. Moreover, the immersive environment will include a cognitive agent to support the system architect with intelligent capabilities: models verification, context-aware recommendation of rules, identification and automation of modelling routines...