AI agents for buildings

The increasing integration of renewable, volatile energy sources calls for advanced plant and operational strategies that can make optimal use of flexibility options on the consumer side. In recent years, innovative building management solutions have been developed for this purpose, which have already demonstrated a CO2 reduction potential of up to 85%. However, as buildings must meet individual requirements and mass production is not possible, the widespread implementation of climate-friendly system solutions requires considerable human resources. This hinders the urgently needed further expansion of sustainable energy infrastructures.

In this context, multimodal AI technologies promise significant progress. They offer the possibility of providing data and information efficiently in a machine-readable format, thereby drastically reducing the manual engineering effort. Furthermore, the integration of AI technologies facilitates the personalisation and customisation of complex systems, improves the prediction and prevention of faults, and supports informed decision-making. In addition, they promote natural interaction between people and technical systems, ensuring more efficient planning, construction and operational management of buildings. This leads to significant scalability and cost benefits.

To capitalise on these benefits, the project is developing AI agents with multimodal capabilities for the holistic optimisation of building services installations and systems. Through the integration of retrieval-augmented generation, sensor data fusion, code interpreters and the embedding of external client-side tools, the AI agents can access domain-specific knowledge databases, take physical principles into account and effectively solve optimisation tasks. Furthermore, the chosen agent-based approach offers decisive advantages: agents can handle specific tasks more efficiently, thereby reducing computational effort and increasing the reliability of the results. At the same time, the interaction of multiple agents significantly enhances the performance and quality of the AI systems. Furthermore, their specialised capabilities improve interdisciplinary and cross-trade collaboration, which supports the integrated planning, construction and operational management of buildings. In the long term, agent-based technologies and methods can thus make a significant contribution to addressing the challenges of climate change in the building sector.

To ensure industry-relevant research, an integrated development environment is being created in which the AI agents can be developed, tested and evaluated for numerous practical applications. This environment comprises both advanced simulation methods and a specially equipped test building that enables real-world testing conditions.