Hybrid vision/force control at corners in planar robotic-contour following

The accuracy and execution speed of a force controlled contour-following task is limited if the shape of the work-piece is unknown. This is even more true when the workpiece contour contains corners. This paper shows how a hybrid vision/force control approach at corners in planar-contour following results in a more accurate and faster task execution. The vision system is used to measure online the contour and to watch out for corners. The edge is correctly located by compensating the compliance of the tool/camera setup which affects the contour measurement. A simple corner-detection algorithm is presented. Once a corner is detected, the finite-state controller is activated to take the corner in the best conditions. Experimental results are presented to validate the approach.