Toward intelligent safety assurance in construction human-robot collaboration

Construction human-robot collaboration (HRC) introduces physical, ergonomic, and cognitive risks that conventional isolated-based safeguards cannot adequately manage in dynamic, unstructured worksites and requires more adaptive forms of safety assurance. Although diverse robotics-related approaches have been developed, how they function as interconnected, active safety-assurance mechanisms for construction HRC remains insufficiently synthesised. This paper systematically reviews 215 studies published from 2011 to 2025 through PRISMA-guided screening, bibliometric analysis and thematic synthesis. Five connected technological directions are identified: robot design optimisation, context-aware perception, human-centric understanding and forecasting, autonomous robotic adaptation, and safety management. The synthesis shows a transition from passive risk avoidance toward perception-driven, prediction, and adaptive safety assurance, but 54% of studies remain limited to controlled laboratory validation, and few integrate uncertainty across sensing, prediction, control and supervision. Accordingly, a closed-loop framework is proposed to guide the integration and construction-realistic validation of verifiable, uncertainty-aware HRC safety systems.