ImmersiProtor: A Collaborative Mixed-Prototype Tool Integrating Spatial Augmented Reality and Component-layered Generation

Paper Title

ImmersiProtor: A Collaborative Mixed-Prototype Tool Integrating Spatial Augmented Reality and Component-layered Generation

Publication Info

• Topic area: Integration of SAR and GAI for collaborative conceptual design.
• Keywords: Spatial Augmented Reality, Generative Artificial Intelligence, Co-design, Mixed-prototype, Component-layered generation, Immersive collaboration, Design iteration, Human-computer interaction, Prototyping tools, Multidisciplinary teamwork.

Background and Problem

• Problem / challenge: Traditional prototyping methods face a trade-off between physical prototypes' tangibility and virtual prototypes' visual fidelity. Existing SAR systems are limited to passive displays and lack creative adaptability, while GAI tools often lack controllability and fail to support fine-grained collaboration.
• Significance: Addressing these limitations can enhance the efficiency, creativity, and collaboration of multidisciplinary design teams, especially during the critical conceptual design phase.
• Motivation and related work: Prior research has explored mixed-prototype methods and GAI for prototyping, but gaps remain in integrating SAR and GAI for real-time, collaborative, and controllable design workflows. This paper builds on these foundations to propose a novel interaction paradigm.

Solution

• Proposed approach: ImmersiProtor, a tool integrating multi-view SAR and component-layered GAI for collaborative conceptual design.
• Novelty:
  1. Introduction of a component-layered generation paradigm to enhance GAI controllability and support fine-grained iteration.
  2. Development of a multi-view SAR system for immersive, consistent, and multi-perspective design evaluation.
  3. Creation of a collaborative design system that balances independent creativity with shared team resources.
• Procedure and key techniques:
  1. Designers create physical prototypes and capture them using cameras.
  2. ImmersiProtor uses global and local prompts to generate high-fidelity, component-layered digital designs via GAI.
  3. Multi-view SAR projects consistent renderings onto the physical prototype from four perspectives.
  4. A web-based platform supports independent and collaborative component creation, iteration, and evaluation.

Results

• Concrete findings:
  • ImmersiProtor achieved an average SUS score of 84.44, classified as "Excellent."
  • NASA-TLX results showed reduced mental demand (U = 10.5, p < 0.01) but increased physical demand (U = 266.0, p < 0.01).
  • CSI score of 87.93 (SD = 2.33) indicated significant support for creativity, especially in expressiveness (U = 274.0, p < 0.01) and collaboration (t = 2.09, p = 0.04).
• Advantage over baselines:
  • Improved collaboration and awareness (U = 228.5, p = 0.034) and shared understanding and consensus (U = 239.0, p = 0.014) compared to the baseline tool (Vizcom).
  • Enhanced creativity support and immersive collaboration compared to screen-based workflows.
• Experiments / evaluation:
  • A user study with 36 participants (18 experimental, 18 control) tasked with designing intelligent delivery vehicles.
  • Measurements included SUS, NASA-TLX, CSI, and a custom collaboration questionnaire.
  • Behavioral analysis identified four creation modes: physical editing, physical combination, component iteration, and component combination.
• Limitations and future work:
  • Limited evaluation across diverse design contexts and tasks.
  • Need for larger-scale studies involving non-design stakeholders and real-world workflows.
  • Challenges with complex structures, intricate appearances, and balancing fidelity with ambiguity.

Summary

ImmersiProtor integrates multi-view SAR and component-layered GAI to address the limitations of traditional and mixed-prototype workflows. It enhances creativity, collaboration, and controllability by allowing designers to iteratively create and evaluate physical and digital prototypes in a shared, immersive environment. Empirical results demonstrate its usability, creativity support, and collaborative advantages over baseline tools. Future work should explore its application across diverse contexts, refine its interaction mechanisms, and balance fidelity with early-stage design ambiguity.