Real-time optimization for a Digital Twin of a robotic cell with human operators

In this paper, we develop a Digital Twin (DT) for a cell with a robotic arm serving seven stations and assisted by a human operator. The lack of reliable predictions of human behaviours requires constant monitoring of the cell in order not to affect the degree of context-awareness, autonomy, and adaptability of a DT. This monitoring is carried out through a real-time simulation and optimization algorithm, which continuously produces near-optimal decisions for machines and efficient recommendations for human operators. The proposed approach is tested in a large computational campaign based on real-world data. The results show that the behaviour of the operator scarcely affects the performance of the cell at least in the considered settings.