An Autonomous Procedure for Scoop-Grasp Exploiting Hand Reconfigurability

Grasping in unstructured environments requires highly adaptable and versatile hands together with strategies to exploit their features to get robust grasps. This thesis presents a method to grasp objects using the novel Soft Scoop Gripper (SSG). The main features of this hand are represented by the scoop and by the reconfigurability of its fingers. The first acts as an embodied environmental constraint (EC), it allows to facilitate the grasp and improves its robustness while also reducing the planning effort. The second allows adopting a single strategy for grasping objects of different sizes. The thesis proposes three fingers’ configurations that are designed to deal with various objects. These configurations are the results of a preliminary grasp analysis of 12 initial settings. The considered grasp strategy is the so-called scoop grasp, where the scoop adapts its orientation to the surface where it slides, while the soft fingers establish reliable contact with the object. The hand positioning for grasping the object is implemented using a recently proposed functional model for soft hands, the closure signature. Besides, some criteria are defined to discriminate the optimal fingers’ setting to approach a given object exploiting the position of the closure application point. Given the object to be grasped, the proposed grasp planner chooses the best configuration of the fingers depending on object geometry and orientation, and then suitably aligns the hand to it. Whereas, a hybrid force-velocity controller is implemented to execute the sliding motion on the surface. The results of 250 grasping trials with 13 different objects using the SSG attached to a robot arm confirm the validity of the criteria used for choosing the best fingers’ setting. Besides, it was shown that the deliberate exploitation of environmental (surface on which the object lies) and embedded (scoop) constraints leads to robust grasps and that fingers’ reconfigurability allows using one strategy (scoop grasp) for several different objects.