Research on Online Positioning Method for Horizontal Directional Drilling Bits Based on Transformer-UKF Algorithm

With the acceleration of urbanization, horizontal directional drilling (HDD) technology has been increasingly applied in underground engineering construction, where the accuracy of drill bit positioning directly impacts construction quality and efficiency. To address this, this study investigates Kalman filter-based algorithms. By integrating Transformer’s temporal model capability with the probabilistic estimation advantages of the Unscented Kalman filter (UKF), the proposed fusion algorithm achieves significant performance improvements: For single-point positioning, it enhances localization accuracy by 90.48% compared to the traditional EKF-STF, 61.11% over the improved AEKF, and 84.95% over the improved ESKF. For dynamic positioning, it improves performance by 16.99% over AEKF and 2.33% over EKF-STF, effectively reducing errors caused by noise and model inaccuracies while enhancing positioning accuracy and robustness. Simulation results confirm that the fused algorithm delivers higher precision in both online single-point and dynamic positioning, significantly improving the accuracy and efficiency of real-time drill bit localization.