Design and grasp planning of a reconfigurable hybrid soft gripper using SINDy and virtual object representation

While grippers based on soft actuators have gained a lot of attention in recent years due to their flexibility and adaptability, they face significant limitations in terms of dexterity and precise control. To contribute to overcoming these limitations, this article proposes a fully 3D-printable, reconfigurable hybrid soft gripper. The proposed gripper is composed of three pneumatically actuated soft fingers that enable adaptation and safe interaction with objects in the environment, and a gear actuation mechanism, which is responsible for increasing dexterity. However, controlling the dexterity of a reconfigurable soft gripper like this requires finding a model that describes the nonlinearities of soft actuators and enables the planning of a firm and reliable grip. The present article also addresses this challenge by utilizing the SINDy (Sparse Identification of Nonlinear Dynamics) algorithm to model, based on a minimal amount of experimental data, the proposed reconfigurable hybrid soft gripper. Subsequently, specialized grasp planning is developed for the proposed soft gripper, which involves the creation of a library of virtual objects for the analysis and identification of contact surface grip points. The model and identified grip points are then input into a grasping solver, which provides a viable energy-based gripper positioning and orientation solution to ensure maximum contact with the object while applying minimal pressure. Several experimental tests were carried out to validate the approach, assessing manipulation dexterity and the ability to achieve a firm grasp on objects. The results demonstrated an 85% success rate in grasping quasi-cylindrical fruits and vegetables, and a 70% success rate for spherical objects.