Collaborative Research: Elements: AI-Agentic Digital Twin for Collaborative Building Auditing and Retrofit Simulation

Many buildings in the United States are aging, energy inefficient, and increasingly difficult to assess and retrofit at scale. In the building energy auditing and retrofitting domain, researchers and practitioners are using technologies such as drones, thermal imaging, artificial intelligence, and building energy modeling to detect envelope problems, evaluate retrofit options, and support decision-making. However, these tools often remain fragmented across data collection, analysis, simulation, reporting, and expert interpretation, making it difficult for the building research community to create reusable workflows, compare methods, validate results, and translate technical advances into practical retrofit-supporting tools. This project will develop a new cyberinfrastructure platform that connects building data, domain knowledge, simulation tools, artificial intelligence, and interactive digital twin environments to support more transparent, reproducible, and evidence-based residential energy retrofit analysis and decision-making. The project serves the national interest by advancing the science of building energy auditing and retrofitting, supporting more reliable and scalable research infrastructure, and contributing to improved energy efficiency, housing performance, and resilience in the built environment. This project will develop REACT, Resilient Energy Assistance through Collaborative Twinning, a cyberinfrastructure framework that integrates artificial intelligence-assisted reasoning with interactive digital twin environments to support collaborative home energy auditing, retrofit design, and multi-stakeholder decision-making. REACT will provide the building energy research community with intelligent and responsive digital twin interfaces for organizing heterogeneous building data, testing retrofit strategies, examining decision processes, and grounding research workflows in realistic operational settings through collaboration with community agencies and practitioners. The project will integrate three core components: 1) digital twin functional modules for managing building data, invoking analytics, and supporting simulation-based retrofit modeling; 2) a knowledge-driven agent that extracts information from audit reports and domain guidebooks, applies retrieval-augmented generation, and performs context-aware reasoning; and 3) a human-digital twin interaction platform that enables interactive retrofit design, real-time feedback, co-learning, and informed analysis among multi-domain researchers and practitioners. Together, these components will form a scalable and participatory cyberinfrastructure platform that helps the building energy auditing and retrofitting community use automation technologies more effectively and develop reproducible, evidence-based workflows for resilient and energy-efficient housing research. The validated REACT framework will be evaluated through collaboration with community action agencies and technology partners. Dissemination through research, technical, and professional networks will include open-source modules, application programming interfaces, trained agents, documentation, and reusable user interfaces. This award by the Office of Advanced Cyberinfrastructure is jointly supported by the Section for Civil, Mechanical, and Manufacturing Innovation in the Directorate for Engineering. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria. NSF Award ID: 2609219 | Program: 01002627DB NSF RESEARCH & RELATED ACTIVIT | Principal Investigator: Kaiwen Chen | Institution: University of Alabama Tuscaloosa, TUSCALOOSA, AL | Award Amount: $299,998 View on NSF Award Search: https://www.nsf.gov/awardsearch/show-award/?AWD_ID=2609219 View on Research.gov: https://www.research.gov/awardapi-service/v1/awards/2609219.html