Annual Journal of IIE Hong Kong

This project addresses the task of Automatic Generation of Gripping Positions for assembly operations. It presents an algorithm to find the gripping faces for a two jaw gripper based on features of the object and the geometry of the object and assembly. The geometric data of the object is obtained from a solid modeler(IDEAS). The geometric data of the object such as faces, edges and vertices described with respect to object coordinate system are obtained from the Solid Modeler. Grasp positions are postulated and then validated or rejected on the basis of this information. Finally the output is shown in the Solid Modeler through a program file. Introduction Task Level robot programming is use d as method to reduce the complexity of robot programming. For example,, rather than instructing the robot which faces to grasp, where to go and close the hand, if the part to be picked is given in the form of solid model, the algorithm automatically determines the gripping faces. Such a procedure would be simpler to use to generate robot programs. This project deals with the important component of the task planning system for automated assembly-the determination of grip positions that a robot may use to pick a part. Literature Review Gatrell [31 proposed a Grasp synthesis utilizing polygons, edges, vertexes. In his algorithm he uses Modified Extended Gaussian Images to compute the pair of graspable entities. Lozano Perez [11 finds out the pair of graspable faces based on reachability and area of contact. In the model he also considers the collision of the object with other objects in the environment and the collision between the gripper and the object surfaces. Lozano Perez et. al. [131 described approaches to determine grasps based on the interaction between the choice of the grasp and the choice of the paths to reach the grasp. The choice of grasp configurations based on considerations of object geometry [11, stability or uncertainty reduction [141. Lozano Perez[ll also determined the potential collisions of the gripper and the neighboring objects in the final grasp configuration, initial grasp configuration, potential collision of the gripper and the neighboring objects in the initial and final configuration. A. C. Woo et. al. [41 in addition to determining the graspable surfaces, considers the factors such as resilience to slippage and twisting in the gripper due to external forces and torques. Aditi Dubey et. al. [8] considers geometric relationships of the object with other components in the assembly in determining the grasping surface. Pierce's Gripes(Purdue Engineering Research Center's Grip Evaluation System)[5l is used to determine the grasping surfaces using a three jaw finger, based on some heuristics and exact coordinates are determined using numerical solution of optimization problems. D.T. Pham et. al. [61 describes a prototype knowledge based system based on component to be gripped, task to be performed, the work environment, the robot and the gripper. Assumptions The following simplifying assumptions are made in the algorithm implemented. 1. All the objects are modelled as a polyhedra, including non convex polyhedra. 2. The robot grasps object using a parallel jaw grippers. 3. At least one object face is involved in the grasp thus there is a well defined grasp plane parallel to this face. 4. A grasp is performed via a plane-to-plane contact. Outline of the Approach The approach to planning the grasp proceeds as follows. Parallelism Determine all possible grasp surfaces that can be grasped using a parallel jaw gripper. Definition: Two planar faces are parallel if the normal vectors of the two faces form antipodal points in a Gaussian sphere or two normal vectors are opposite to each other. Local Accessibility This analysis is aimed at eliminating those pairs determined from step 1 which cannot provide contact with a jaw of the gripper. It is equivalent to finding out whether there are any convex edge on the surface being grasped and if the length of the convex edge can accommodate the gripper. As shown in Figure 1, though surface A and B form an antipodal point, surface A is not accessible by the gripper, so such surfaces will be eliminated. FIGURE 1: Example part shows surface A is locally inaccessible 1994-95 Annual Issue of IIE (HK) 13 Definition:The orientation vector of the edge ai belonging to face F is: