Prior-knowledge assisted fast 3D map building of structured environments for steel bridge maintenance

Abstract

Practical application of a robot in a structured, yet unknown environment, such as in bridge maintenance, requires the robot to quickly generate an accurate map of the surfaces in the environment. A consistent and complete map is fundamental to achieving reliable and robust operation. In a real-world and field application, sensor noise and insufficient exploration oftentimes result in an incomplete map. This paper presents a robust environment mapping approach using prior knowledge in combination with a single depth camera mounted on the end-effector of a robotic manipulator. The approach has been successfully implemented in an industrial setting for the purpose of steel bridge maintenance. A prototype robot, which includes the presented map building approach in its software package, has recently been delivered to industry.