High-speed Robot Motion Control under Visual Guidance

The paper describes a vision based method and implementing procedure allowing for collision-free, on-the-fly grasping of objects travelling on conveyor belts. The ensemble conveyor belt + actuator + sensor is configured as a m ¿ 3 -axis Cartesian robot, leading to a problem of cooperation between multiple robot manipulators subject to the multitasking control of a computer. The collision-free grasping of recognised and located objects is checked at run time by projecting the fingerprints of the gripper onto the image plane, and estimating whether they "cover" only background pixels. A structured programming environment is used for exemplifying the multi tasking control of horizontally articulated (SCARA) robots tracking moving objects for collision-free grasping, under guidance vision; experimental results are also reported.