Researchers at Chongqing University have developed an adaptive Kalman filter localization algorithm for wheeled robots operating in underground drainage networks with diameters of 400–600 mm. The method fuses low-cost wheel encoders, optical flow sensors, and a six-axis inertial measurement unit to estimate the robot's trajectory in GPS-denied environments, addressing the challenges of wheel slippage and error accumulation common in confined, slippery pipe conditions.
Key takeaways
- An outlier detection mechanism with a set threshold corrects wheel encoder data, mitigating the impact of wheel slippage.
- The Taguchi method tunes the process noise covariance matrix, achieving efficiency five times that of trial-and-error approaches.
- Residuals, a forgetting factor, and parameter thresholds adaptively adjust the measurement noise covariance matrix in real time.
- Physical experiments show the algorithm suppresses longitudinal slip errors, delivering positioning accuracy 3.26 times that of an extended Kalman filter-based method.
This work provides a practical solution for autonomous inspection and maintenance robots in urban sewer systems, where reliable localization is critical for documenting pipe defects and avoiding robot loss. The adaptive filtering approach enhances robustness in slippery, feature-sparse environments, offering a cost-effective alternative to high-end sensors for industrial pipe inspection applications.
Source: 《机器人》期刊 (robot.sia.cn) · Published 2026-01-13 · “面向地下排水管网的轮式机器人自适应卡尔曼滤波定位算法”