An Integrated Vision-Force System for Peg-in-hole Assembly Operations

In this article the integration of an artificial vision system and a 6-component force-torque sensor for high accuracy insertions is proposed. The former has the advantage of intervening before the contact of the two parts and providing an initial correction to allow the force sensor action; the latter is charged of completing the operation through the control of a neural network. An experimental facility has been developed for performance assessment. Two monochromatic CCD cameras are positioned with the optical axes perpendicular to each other, outside the working area in order not to interfere with the assembly operations. During tests it has been shown that the artificial vision system yields a fast, reliable and accurate measurement, by direct comparison with a co-ordinate-measuring machine. Results show that the use of an integrated vision-force sensor represents a complete solution to the peg-in-hole problem. Furthermore the 3-D localisation algorithm implemented on the vision system and the problems concerned with the system engineering are dealt with.