High-Precision 3D Mapping In Tunnels Based on Vibration Magnitude-Adaptive Kalman Filtering

The 3D digital map of tunnel walls is crucial for the precise control of automated shotcrete operations in tunnels. The accuracy of pose estimation for point cloud sensors directly affects the precision of the 3D digital map, especially in dynamic vibration environments where time-varying noise interference poses significant challenges. This study proposes a vibration-adaptive Kalman filtering (VAKF) framework that integrates total station, inertial measurement unit, and motion constraints for high-precision pose measurement of the tunnel boring machine shotcrete arm. The method employs a two-level fusion strategy: attitude-level fusion dynamically adjusts the noise covariance of the inertial measurement unit and total station based on vibration magnitude to correct the angle; position-level fusion embeds circular motion constraints to suppress cumulative integration errors. Experiments conducted in a simulated tunnel environment demonstrate that the proposed method achieved an RMSE of 34.87 mm under the working conditions of an arm length of 1m and a movement angular velocity of 0.1 rad/s, outperforming the measurement accuracy of static total station measurement, IMU integration measurement, and Sage-Husa adaptive filtering. Additionally, this method maintains robust performance across different arm lengths and movement speeds, with the RMSE consistently remaining below 41.539 mm. This study addresses the limitations of existing sensor fusion methods in vibration scenarios, providing a practical solution for real-time 3D mapping in automated tunnel construction.