Abstract: An autonomous attitude control system, utilizing a combination of artificial intelligence and traditional control theory, is proposed to enhance the quality of tunnels formed and minimize environmental disruption during the excavation process. To achieve this, the control system is divided into two layers: planning and execution. The planning layer employs a collection experience modal decomposition method to process the highfrequency signal in the original data in a stable manner. After training the historical and realtime data, a shield motion model is established to plan the control trajectory based on the design axis. In the execution layer, a position controller model is built using a neural network, and the advancing parameters are adjusted according to realtime feedback, allowing for the prediction and autonomous control of the shield attitude. The system was implemented in the ZhiYu shield excavation control system, utilized in the leftline of the KeqiaoDiyang section of the HangzhouShaoxing metro tunnel project. The results demonstrate that the proposed control system outperforms the manual control method in terms of stability and prediction accuracy. The autonomous attitude control system presents a promising solution to enhance tunnel excavation processes and reduce environmental disturbance.

Key words: shield, autonomous attitude control system, datadriven, data noise reduction, automatic control