ARCHIVES

Human–robot collaboration (HRC) has become a cornerstone of smart manufacturing in Industry 4.0, enabling enhanced productivity, flexibility, and human-centric automation. However, ensuring safe and efficient interaction between humans and collaborative robots remains a significant challenge due to the limitations of conventional reactive safety mechanisms, which fail to anticipate dynamic human behavior. This paper presents a comprehensive review of artificial intelligence–driven approaches for intent prediction and adaptive safety in collaborative robotic systems. It systematically analyzes recent advancements in multimodal sensing, machine learning, and deep learning techniques used for human activity recognition and predictive modeling in HRC environments. A structured taxonomy of intent prediction methods is developed, covering vision-based models, motion analysis approaches, and sensor fusion techniques, along with their respective strengths and limitations. Building on the identified research gaps, the study proposes a conceptual AI-driven adaptive safety framework that integrates multimodal sensing with predictive analytics to enable proactive human–robot interaction. The framework incorporates deep learning architectures for real-time intent recognition and dynamic control strategies for trajectory adjustment, speed modulation, and task coordination. The paper highlights that despite significant progress in perception and recognition, limited attention has been given to real-time predictive safety adaptation in industrial settings. By combining systematic review, critical analysis, and a forward-looking framework, this work provides a comprehensive foundation for the development of next-generation intelligent collaborative robotic systems