In context
Wall-climbing robots are increasingly used for welding, grinding, and inspection on large steel structures such as hydropower penstocks, storage tanks, and ship hulls. However, these environments often lack visual texture, are enclosed, and suffer from strong magnetic interference, making reliable autonomous localization challenging. Traditional methods like GPS, UWB, or lidar-based SLAM struggle in such conditions, and encoder/IMU dead-reckoning accumulates errors over time.
What was reported
Researchers at Tsinghua University proposed an ArUco-assisted autonomous localization method (A-IEF) that uses a fisheye camera to observe a single ground-fixed ArUco marker. The method integrates IMU, wheel encoders, and the camera through an error-state extended Kalman filter (ES-EKF). It selects key frames based on ArUco position, optimizes pose via reprojection error minimization, and corrects heading and position using encoder displacement errors and ArUco corner reprojection errors.
Experiments on large steel components showed position estimation errors within 0.06 m and heading angle errors within 3.7°. Compared to ArUco-rectified and dead-reckoning methods, the proposed approach reduced position error by about 47% and heading error by about 68%, and it remained functional in low-light conditions.
The system uses a 180° fisheye camera to capture the marker from a wide area, avoiding the need for dense marker placement. The ArUco code is mounted on a ground-fixed LED lightbox, and the robot's attitude is estimated using accelerometer data to establish correspondence between detected corners and actual marker corners.
Why it mattered
This work addresses a critical gap in autonomous navigation for climbing robots in industrial settings, offering a cost-effective and robust alternative to expensive tracking systems or dense marker arrays. By enabling reliable localization on large steel structures, it paves the way for fully autonomous inspection and maintenance operations, reducing the need for manual remote control and improving efficiency and safety.
“Compared with the ArUco-rectified method and the dead-reckoning method, the proposed method reduces the position error by 47% and the heading angle error by 68%, and can implement localization in weak-light environments.”
Source: 《机器人》期刊 (robot.sia.cn) · Published 2024-02-04 · “ArUco辅助的爬壁机器人自主定位方法”