Leveraging Large Language Models and Robotics for Sustainable Construction Project Management

Abstract

The increasing complexity of construction projects has intensified the need for intelligent systems capable of supporting planning, monitoring, decision-making, and operational control while simultaneously addressing sustainability objectives. Large Language Models (LLMs) and robotics represent two complementary technological capabilities: LLMs can support information interpretation, reasoning, communication, and decision assistance, whereas robotic systems can translate digitally generated decisions into physical construction activities. This review-oriented research paper develops a conceptual framework for integrating these technologies into sustainable construction project management. The methodological foundation is derived exclusively from the supplied literature, which primarily examines neural-network-based classification, image processing, and machine-learning approaches for medical-image analysis. Although these references do not directly investigate construction management, their treatment of automated classification, feature interpretation, image-based recognition, and machine-learning decision processes provides transferable methodological principles for intelligent construction monitoring. The proposed framework therefore positions LLMs as a cognitive coordination layer and robotics as an execution and sensing layer within a sustainability-oriented project-management architecture. The analysis identifies potential benefits in automated progress interpretation, resource coordination, anomaly identification, safety-oriented monitoring, and sustainability decision support. It also identifies critical limitations, including domain-transfer uncertainty, data quality, model reliability, explainability, interoperability, and the danger of treating domain-specific machine-learning evidence as direct validation for construction applications. The study contributes a theoretically grounded integration model while emphasizing that empirical construction datasets and field experiments are required before the framework can be considered operationally validated.