Grasping of Deformable Objects Applied to Organic Produce

Grasping mechanisms (grippers) are used for a wide range of applications including fixturing arrangements, industrial, agricultural, and service robotics for medical and home use. The gripper must immobilize the object it is manipulating , while applying the minimal necessary grasping force, in order to prevent the grasped object's bruising. Performing grasp analysis and synthesis, requires the development of a grasp model. The grasp model includes kinematic, dynamic and contact models, where the last define the coupling between the contact forces and their deformations [4]. All grasp models greatly depend on the system's parameters, such as the number of contact points, the object's geometry, the gripper's structure, the desired object's manipulation, and the mechanical properties of the object and gripper [3]. Contact models were developed in order to assess and predict the reaction forces and local deformations. They vary from considering rigid components and sharp fingertips with no friction at the contact point, to soft finger models – assuming compliant components, which imply a contact area with a force and torque distribution [1,5]. Note that the contact's location is referred as a contact point, although geometrically it is a contact patch in some of the discussed compliant contact models. The goal of this work is to provide the set of grasp forces and torques that can be applied to an object, without losing equilibrium and stability constraints. The grasp planner, provides an optimal set of positions for n contact points, which are optimal in the sense of minimizing contact force and torque at each contact point. The planning is independent in the selection of a specific gripper, as oppose to more traditional techniques [2]. Under the constraints of maintaining grasp's stability and not bruising the object, two grasp – quality criteria are currently in consideration: 1) The minimal required grasping force and torque. 2) The maximal allowed grasping force and torque. In order to achieve this goal, the following assumptions and definitions are made: consider a virtual gripper, whose n hemisphere fingertips can be positioned at every point in space, and actuated at any desired direction. The fingertips and the object are considered as compliant (quasi-rigid) [1,6]. The object's B geometry and mechanical properties are known, and it may be subjected to a known set of external forces and torques (loads) ˜ F. The object can be grasped