Hybrid Vision – Language & Knowledge-Graph Framework for Circular Construction Waste
DOI:
https://doi.org/10.66408/dfbi.2026.26381Keywords:
Construction Waste Management, Circular Economy, Knowledge Graphs, Vision–Language Models, Logical Reasoning, Sustainable ConstructionAbstract
Construction activities generate significant amounts of waste, posing serious challenges to environmental sustainability and resource efficiency. Circular economy strategies, particularly the 5R principles (Reduce, Reuse, Recycle, Recover, and Dispose), provide a structured approach for improving construction waste management; however, effective implementation on construction sites remains difficult due to visual complexity, mixed materials, and limited decision-support tools. Existing waste monitoring methods, including vision-based and data-driven approaches, often lack reliability, interpretability, and alignment with circular-economy-oriented decision-making. This paper proposes a hybrid conceptual framework that integrates Vision–Language Models (VLMs) with a knowledge-graph-based logical reasoning engine to support circular construction waste management. In the proposed approach, VLMs interpret construction-site images and extract semantic information related to waste materials and contextual conditions. The extracted information is structured within a knowledge graph, where a logical reasoning engine applies rule-based inference aligned with the 5R principles to ensure consistent, transparent, and explainable waste classification and handling decisions. By embedding circular economic logic as a dedicated decision layer, the framework shifts the role of AI systems from simple waste detection toward sustainability-driven decision support. The paper reviews related work on construction waste management, vision-based waste detection, and knowledge-driven reasoning systems. It identifies key challenges and opportunities associated with intelligent waste monitoring on construction sites. The proposed framework demonstrates how combining multimodal perception, structured knowledge representation, logical reasoning, and human-in-the-loop feedback can enhance reliability, transparency, and sustainability. This work lays the foundation for developing practical, interpretable, and circular economy-oriented construction waste management systems.
Downloads
Published
Conference Proceedings Volume
Section
License
Copyright (c) 2026 Mina Sadat Orooje, Nazly Atta, Cinzia Maria Luisa Talamo, Fulvio Re Cecconi (Author)

This work is licensed under a Creative Commons Attribution 4.0 International License.