Large Language Models for Intelligent Decision-Making in Robotics-Enabled Sustainable Construction

Abstract

The integration of Large Language Models (LLMs) with robotics-enabled construction represents an emerging socio-technical approach to intelligent decision-making in complex and sustainability-oriented construction environments. This paper develops a conceptual research framework for understanding how language-based artificial intelligence can support robotic systems in interpreting operational information, coordinating activities, prioritizing construction decisions, and responding to uncertainty. Because the supplied literature primarily addresses data-driven prediction, customer-behavior modeling, class imbalance, feature selection, and social-network effects rather than construction robotics or LLMs directly, the study adopts a theory-transfer methodology. The analytical foundation is derived exclusively from the seven supplied references and extends their principles of predictive modeling, feature selection, imbalance management, survival-oriented reasoning, and relational analysis toward construction decision environments. The proposed framework conceptualizes LLM-enabled robotic decision-making as a layered process comprising data acquisition, contextual interpretation, predictive reasoning, decision generation, robotic execution, and feedback-based adaptation. The analysis indicates that intelligent construction systems require more than language generation: they require reliable data representation, uncertainty management, contextual prioritization, and mechanisms for translating analytical outputs into operational actions. The study identifies potential benefits in resource allocation, safety-oriented planning, equipment coordination, workflow adaptation, and sustainability management while emphasizing limitations involving domain transfer, data quality, interpretability, imbalance, and decision accountability. The paper contributes a theoretically grounded framework for future empirical investigation of LLM-driven decision intelligence in robotics-enabled sustainable construction.