Abstract: Real-time and high-precision positioning of inspection robots remains a crucial technical challenge for the operation, inspection, and maintenance of long-distance water conveyance tunnels, particularly under water-filled conditions. Conventional localization methods such as inertial navigation, odometry, and vision-based systems often accumulate errors over long distances and are severely affected by environmental noise, signal attenuation, and the geometric constraints of tunnel environments. To address these shortcomings, this study proposes a novel acoustic positioning method based on the Differential Threshold Method (DTM), aiming to enhance robustness, accuracy, and reliability for robot localization in complex tunnel settings. The proposed approach employs a system of acoustic positioning base stations strategically deployed along the tunnel. Each base station can receive acoustic signals emitted by a mobile terminal integrated into the inspection robot. The algorithm adopts a two-stage time delay estimation process. First, the generalized cross-correlation (GCC) technique is used to obtain a coarse estimate of the time delay differences between signals captured at various base stations. Then, the Differential Threshold Method is applied for fine-grained estimation of these differences. These refined estimates enable high-accuracy coordinate calculations of the mobile terminal using time difference of arrival (TDOA) principles. To thoroughly evaluate the effectiveness and practicality of the proposed system, both continuous dynamic laboratory-scale experiments and full-scale field tests were conducted. During these trials, varying ambient acoustic noise levels and diverse tunnel configurations were simulated to replicate realistic engineering conditions. The comprehensive experimental assessment included measurements of positioning accuracy, robustness to noise and signal attenuation, and system adaptability to changing tunnel geometries. Experimental results demonstrate that the DTM-based acoustic positioning technique consistently achieves positioning accuracy better than 0.10 meters along the tunnel’s axial direction, thereby meeting and surpassing the stringent engineering standards required for water conveyance tunnel inspection robots. The system also demonstrates substantial robustness against environmental disturbances, maintaining stable performance under adverse noise conditions and signal attenuation. Comparative evaluations show that, compared with traditional techniques, the proposed approach delivers notable improvements in localization precision and operational reliability. Moreover, the system’s real-time processing capability enables around-the-clock inspection and maintenance operations without requiring water flow interruption—an essential requirement for large-scale water infrastructure where supply discontinuity is costly or impractical. The modular nature of the hardware and software design allows easy scaling and customization for varying tunnel lengths and diameters, ensuring broad applicability across diverse engineering projects. In summary, this research presents a DTM-based acoustic positioning technology that significantly advances the state of the art in inspection robot localization under water-filled tunnel conditions. The methodology integrates generalized cross-correlation and differential threshold estimation to deliver enhanced accuracy and reliability, as validated by extensive experimental data. This work provides not only a technical solution that meets current industrial demands for precision and robustness, but also establishes a solid foundation for future integration with fully autonomous tunnel inspection platforms. Future work will focus on further miniaturization of sensor modules, improvements in power efficiency, and seamless integration of the positioning system with autonomous navigation frameworks. The approach has strong potential for broad adoption in long-term infrastructure maintenance and may be extended to other subsurface or high-interference environments requiring robust, high-precision robotic localization.