A Prior-Depth VIO and Visual Fiducial Fusion Positioning Method for Climbing Robots on Wind Turbine Blades

Climbing robots exhibit extensive motion capabilities and flexible deployment, making them a promising solution for processing wind turbine blades. Achieving large-scale and real-time positioning for climbing robots is a key challenge for automated applications. This paper proposes a practical and cost-effective positioning method. It requires only one external observation camera, one onboard camera, and one inertial measurement unit (IMU). A dual-sensor fusion framework is proposed, which comprises a prior-depth visual-inertial odometry (VIO) and an AprilTag-based visual fiducial positioning module. After rigorous coordinate system alignment, the proposed framework employs Kalman filter to fuse information from VIO and AprilTag-based positioning method, which enable robust global positioning with enhanced accuracy and motion continuity. The proposed method was verified on a large wind turbine blade with a planar projection size of 3 m × 1.5 m. The robot’s total movement length reached 16.28 m, with a maximum drift of 6.0 cm.