Surface reconstruction and path planning for industrial inspection with a climbing robot

Inspection and maintenance tasks in the power industry often require tools to be moved smoothly in direct contact or close proximity over surfaces of curved tube-like structures. This paper addresses the problem of navigating robotic devices, namely a wheeled climbing robot carrying an inspection tool, on surfaces as they typically appear in steam chests of power plants. The surfaces are modeled by triangle meshes. Several state-of-the-art surface reconstruction methods are evaluated, and meshes are constructed for 3D point clouds taken with different laser scanners and for tube structures varying in shape, dimension and surface property. Each mesh represents an embedded graph, and thus directly enables graph search methods to plan paths of triangle strips over the surface. Discrete path planning and continuous robot control are combined to a hybrid system, which steers the robot smoothly along the triangle strip to the target regions on the surface. The navigation approach applies to robotic surface inspection, and is verified in simulation experiments by moving a robot on triangle meshes reconstructed from real data.