A method for efficient virtual-physical synchronization of the digital twin system of an excavation-supporting robot cluster targeting coal pillars between mining faces

ObjectiveVirtual models for excavation-supporting robot clusters targeting coal pillars between mining faces encounter challenges like a large data size and anomalies in data transmission, which lead to poor virtual-physical synchronization. This study proposed a method for efficient virtual-physical synchronization of the digital twin (DT) system of an excavation-supporting robot cluster using 3D model lightweighting and a trajectory prediction and correction model. MethodsFit-controlling vertices were defined, and their collapse cost factor was introduced to improve the quadratic error metric (QEM) algorithm and to constrain the lightweighting process of the 3D model of an assembly while maintaining fits between components. This leads to a decreased data size. A trajectory prediction-correction model was developed for the excavation-supporting robot cluster. Specifically, the movement trajectories of the twin robot cluster were predicted using the self-attention-long short-term memory (LSTM)-based trajectory prediction algorithm, followed by the real-time correction of the predicted trajectories using quadratic interpolation. This helps ensure the spatiotemporal consistency of the synchronization between the virtual model and the physical equipment. Furthermore, a simulation platform was constructed for DT-based efficient virtual-physical synchronization of an excavation-supporting robot cluster. Results and Conclusions The results indicate that the lightweighting process under the constraint of the collapse cost factor of fit-controlling vertices effectively suppressed the geometric error propagation while maintaining the mating surfaces in the assembly roughly unchanged, achieving a data compression ratio of 90%. For the prediction of the movement trajectories within 1.5 s, the self-attention-LSTM-based prediction algorithm yielded the lowest errors. The trajectory prediction-correction method reduced the mean absolute deviation (MAD) of the driving trajectory by 74.28%, effectively ensuring consistent, stable virtual-physical synchronization. The results indicate a maximum virtual-physical synchronization latency of 55.28 ms, an absolute positional error of 1.93 mm, and a relative positional error of 1.07%, suggesting high-accuracy, low-latency virtual-physical synchronization of an excavation-supporting robot cluster. The proposed method provides a new philosophy for enhancing the operational efficiency of the DT system of coal mining equipment.