Localization and Navigation of a Climbing Robot Inside a LPG Spherical Tank Based on Dual-LIDAR Scanning of Weld Beads

Mobile robot localization is a classical problem in robotics and many solutions are discussed. This problem becomes more challenging in environments with few and/or none landmarks and poor illumination conditions. This article presents a novel solution to improve robot localization inside a LPG spherical tank by robot motion of detected weld beads. No external light source and no easily detectable landmarks are required. The weld beads are detected by filtering and processing techniques applied to raw signals from the LIDAR (Light Detection And Ranging) sensors. A specific classification technique—-SVM (Support Vector Machine)—is used to sort data between noises and weld beads. Odometry is determined according to robot motion in relation with the weld beads. The data fusion of this odometry with another measurements is performed through Extended Kalman Filter (EKF) to improve the robot localization. Lastly, this improved position is used as input to the autonomous navigation system, allowing the robot to travel through the entire surface to be inspected.