On-line automatic robot programming: A case study in grasping

Despite much research in the field, the use of automatic robot programming systems is still a long term issue; it is the reason why we think that, right now, we have to use the methodology supplied by this research to develop object level on-line robot programming tools. Moreover, we think that some accessibility problems which automatic robot programming addresses might be solved a better way in a sensor based approach. In this paper, we focus on the problem of accessibility analysis for automatic grasping. The problem is to choose a goal configuration and a safe path for a 3D object moving bi-dimensionally among 3D obstacles. We propose an approach relying on a bidimensional geometric reasoning; our approach consists in discretizing the space swept by the gripper using a set of parallel planes. Then, an accessibility analysis using the Configuration Space approach is performed in each of these planes and local constraints are transformed in constraints on the motion of the whole gripper. In this paper, we describe this approach and the work in progress for using visual data in building a model of the empty space.