Adaptive Vision-Based Control for Robotic Tiling with Uncalibrated Cameras and Limited FOV

Abstract

The traditional manual tiling is labor-intensive and now limited by the shortage of skilled labors and the increasing manpower cost. While a few automatic machines have been developed to alleviate the problems, the autonomous capability of existing systems is relatively low, in the sense that humans usually involve to calibrate the spatial relationship between the robot and the target tile, specify the desired position of each tile, or deal with unforeseen changes (e.g., falling of tile, temporary loss of features). This paper proposes a new adaptive vision-based control scheme for robotic tiling, which enables the robot to automatically pick or re-pick the tile then place it to the desired position in the presence of uncalibrated cameras and limited field of view (FOV). The proposed controller improves the autonomous capability of tiling robots in twofold. First, the unknown spatial relationship due to uncalibrated cameras is estimated online, such that the manual calibration is not required. Second, the temporary loss of features due to the limited FOV is addressed with the Cartesian-space regional feedback, such that the manual assistance is also eliminated. The stability of the closed-loop system is rigorously proved with Lyapunov methods, and experimental results are presented to illustrate the performance of the proposed control scheme.