3D SLAM Applied to an Autonomous Interior Finishing Robot

摘要

The use of robots in construction site is conducive to the standardization and systematization of the construction process. However, application of robot technology in this field is still limited. Autonomous interior finishing is still far from realization. One of the reasons for this status is the lackness in SLAM technologies oriented to application in autonomous interior finishing scenario. The related technology for environmental sensation has not been broken through. In this paper, we propose a SLAM solution specialized for autonomous interior finishing work. This scheme employs an artificial projection system to increase the visual features in the working site. Visual SLAM is employed to build the map of the working site. And a deep learning based defect detection system is integrated in the scheme of Visual SLAM. It contributes to register the type and position of the defects on walls into the built map for robot navigation. With the additional information about the wall defects, automatic execution of finishing work becomes possible. The proposed method is tested in an actual interior finishing site. The experimental results proved its effectiveness in tasking planning.