Development of a Robot System for Applying Putty on Plastered Walls

Applying putty on plastered walls is one process of interior decoration work, it is characterized by high requirements on flatness and smoothness. Although great progress has been made on sensing technology and robotics in recent years, it is still a big challenge to complete this process automatically by robots. The main problem involved is to design a robot system which can automatically perform putty flattening according to the surface status during the process. Each operation of putty flatting needs to consider the specific uneven condition of the target region which is resulted from the previous operation. Generally, it is difficult to detect the exact unevenness status of a plastered wall surface, since the 3D contour measurement to it is beyond the capability of most related sensors. In this paper, a solution based on vision feedback is proposed for realizing automatic putty painting. The proposed method is characterized as an end-to-end artificial neural network which directly derive the necessary robot operation from the captured image on the target process wall region. The proposed method is integrated into an autonomous interior finishing robot. The experiments conducted demonstrate the effectiveness of the proposal.