Predictive robotic contour following using laser-camera-triangulation

In robotic contour following tasks such as sewing, grinding or applying adhesives, the robot moves a tool along a desired path on a workpiece. In order to compensate inaccuracies of the path definition a vision system can be used to guide the robot during task execution. This paper presents an algorithm to adjust the position and orientation of the tool by predictive vision-based control that compensates system delays caused by the robot dynamics and the vision sensor. Our system includes feature validation for suppressing optical disturbances; tool alignment with minimum angular tracking error along unknown contours; and restricting the tool velocity online depending on the applicable angular velocities. Our experimental setup uses an industrial laser-camera-triangulation sensor providing taught features in 3D such as a visible edge along the surface of the workpiece. The paper includes experimental results of robotic contour following using a wavy compliant surface to validate the approach.