EUREXA: End-User Reconfiguration of Environment with eXplainable Augmentation for Generative Fabrication

Paper Title

EUREXA: End-User Reconfiguration of Environment with eXplainable Augmentation for Generative Fabrication

Publication Info

• Topic area: Generative AI for physical object augmentation and fabrication.
• Keywords: generative design, parametric models, accessibility, multimodal input, fabrication, augmentation, RAG, MLLM, OpenSCAD, user-centered design.

Background and Problem

• Problem / challenge: Non-expert users face barriers in identifying latent interaction challenges, articulating design goals, and accessing advanced fabrication solutions, which are often confined to research prototypes or non-parametric designs.
• Significance: Addressing these barriers can democratize access to generative fabrication, enabling users to create personalized, accessible, and sustainable solutions for everyday environments.
• Motivation and related work: Prior works on parametric design and text-to-3D generation have advanced customization but lack contextual fit and accessibility for non-technical users. Foundation models offer potential for bridging this gap by enabling multimodal reasoning and retrieval-augmented generation.

Solution

• Proposed approach: EUREXA, an agentic system that supports end-users through a diagnose–discover–describe workflow to generate fabrication-ready designs from multimodal inputs.
• Novelty:
  1. Dual search across public repositories and research articles for augmentation solutions.
  2. Conversion of non-parametric designs into parametric models with explainable parameters.
  3. CLEAR metrics to evaluate user input ambiguity and system reasoning.
  4. Human-in-the-loop design iteration with dynamic GUI for customization.
• Procedure and key techniques:
  • Diagnose latent interaction challenges using multimodal input (text and images) and scene analysis.
  • Discover solutions via retrieval-augmented generation (RAG) from repositories and research papers.
  • Describe designs with user-centric parameters and provide a dynamic GUI for iterative customization.
  • Generate parametric designs using fallback hierarchies, including remixing existing designs, converting non-parametric meshes, and drafting designs from research descriptions.

Results

• Concrete findings:
  • Object detection recall: 86.9%.
  • Motion type detection precision: 95.2%.
  • Accessibility issue detection recall: 74.78%.
  • Solution discovery success rate: 95% in the full pipeline.
  • User study: Participants successfully generated and customized designs, reporting high usability and intuitive interaction.
• Advantage over baselines:
  • Outperforms text-only and image-only configurations in solution discovery (95% vs. 30–57.5%).
  • Reliable generation of fabrication-ready designs compared to ablated setups.
• Experiments / evaluation:
  • Evaluated on AccessLens and AccessMeta datasets for object detection and issue identification.
  • CLEAR metrics introduced to quantify input ambiguity.
  • Ablation study and user study with 10 participants from diverse technical backgrounds.
• Limitations and future work:
  • Token limits in RAG restrict long prompts and multi-turn context.
  • Parametric generation errors in complex designs.
  • Limited support for large-scale fabrication, multi-material designs, and hardware constraints.
  • Future directions include AR integration, semantically aware parametric reconstruction, and expanded domain scalability.

Summary

EUREXA is an agentic system that enables non-expert users to diagnose interaction challenges, discover augmentation solutions, and describe fabrication-ready designs through multimodal inputs and iterative customization. It leverages multimodal large language models, retrieval-augmented generation, and parametric design generation to support diverse use cases in accessibility, smart home automation, and energy harvesting. Evaluation demonstrates high accuracy in object detection, solution discovery, and user satisfaction, with future work focusing on AR integration, scalability, and enhanced parametric reconstruction. EUREXA aims to democratize access to advanced fabrication solutions and promote research-backed parametric design sharing.