Multi-agent systems for digital twin shop-floor resilient control

Digital twin (DT) shop-floors represent a critical enabler for smart manufacturing. Leveraging the core mechanism of virtual-real interaction, DT shop-floors (DTS) demonstrate exceptional capabilities in remote monitoring, online analysis, dynamic decision-making, and continuous optimization. However, ensuring the consistency and effectiveness of virtual-to-real control in production processes characterized by uncertainties and interaction delays remains a significant challenge. Although DTS resilient control proves to be a practical solution, it is sensitive to simulation speed, problem scale, algorithm complexity, and uncertainty frequency, while lacking the capability to address emerging uncertainties. To address these issues, a multi-agent system-based method for DTS resilient control is proposed. The system integrates 4 M agents with multi-role collaboration, multi-strategy flexible decision-making, multi-scale adaptive configuration, and multi-process simulation and learning. The definition and connotation of the 4 M agents for DTS resilient control is elucidated. The implementation methods of the 4 M agents are further detailed. Finally, a case study on a DTS for chemical fiber production validates the effectiveness and feasibility of the proposed method.