Towards seamless and safe human-robot collaboration in Industry 5.0: advances in human behaviour prediction

In the era of Industry 5.0, human-centered manufacturing systems are gaining prominence, emphasising the integration of the human worker’s ingenuity, flexibility, and dexterity with the robot’s strength, precision, power and speed, which indicates the transition to seamless Human-Robot Collaboration (HRC). However, the close proximity between human workers and robots also raises safety issues, which necessitates the human behaviour prediction to avoid potential collisions, risks and injuries. This review comprehensively explores the advancements in human behaviour prediction methods in HRC, focusing on mechanism-based, data-based, and mechanism and data-fusion-based human behaviour prediction methods. Then, the major characteristics, limitations, and applicability of the approaches are analysed. Finally, the existing challenges and future directions for human behaviour prediction are discussed, and a unified human behaviour prediction framework is proposed to improve the safety, efficiency, and generalisation in human-centric manufacturing systems.