A hybrid positioning method for eye-in-hand industrial robot by using 3D reconstruction and IBVS

This paper addresses the accuracy problem of 3D positioning of a large-environment and visual servo control for a robot in an unstructured scene. The important issue in this application is the pixel measurement error in 3D reconstruction model. Because of large-task space corresponding to each pixel even a small measurement error can cause significant errors in reconstructing 3D scene. On the other hand, IBVS provides the complementary benefits of accuracy up to required criteria, as a result of close loop system and smaller task-space. However, it has problems of stability and convergence in large movements and rotations. In this approach instead of over calibrating or increasing the resolution of the picture in hand which usually takes a lot of time and process without any satisfying result; a hybrid algorithm is developed to compensate for the error in 3D model, using partitioned IBVS. The proposed method uses the 3D model to moves the end-effector near the target, which makes it possible to implement any complex motion-planning or object-recognition, then uses the partitioned IBVS to fix the error caused by 3D reconstruction. The accuracy by this method can be increased to meet the most requirements.