Multimodal Large Language Models in Construction Education for Learning Human–Robot Collaboration: A Narrative Review

The construction industry is increasingly adopting collaborative robotics to address challenges such as skilled labor shortages, safety risks, and productivity inefficiencies. A critical aspect of this transition is human–robot collaboration (HRC), in which humans and robots work together to achieve shared objectives, thereby enhancing operational efficiency and workplace safety. However, current construction education models often fall short in equipping students with the competencies needed for effective HRC. This narrative literature review explores how multimodal large language models (MLLMs) can support learning of HRC in construction education by synthesizing evidence from five relevant domains (i.e., manufacturing, healthcare, robotics, construction, and education) selected based on their shared emphasis on human–machine interaction, adaptive learning, safety-critical decision-making, and task coordination. Through cross-sectoral analysis, the review identifies the capabilities of MLLMs, including multimodal data interpretation, natural language communication, simulation-based learning, and personalized learning and feedback, which can facilitate the learning of HRC in construction. This review provides a conceptual basis for integrating MLLMs into construction education, highlighting their potential to support immersive, personalized, and artificial intelligence-enhanced HRC learning experiences. It also identifies the challenges of applying MLLMs to learning HRC and suggests strategies to address them. This study contributes to knowledge by investigating the adoption of MLLMs in construction education to support HRC learning. It identifies key capabilities of MLLMs that can enhance HRC training and outlines future research directions for their effective implementation in construction education.