Evaluation of Grasp Quality Measures Using Tactile Sensors

This thesis presents the implementation of widely recognized grasp quality measures, typically confined to planning, within a real-world experimental setup. The objective is to validate the theoretical foundations of these metrics in practical applications. The hardware components used in the setup designed for evaluating grasp quality measures include the robotic Allegro Hand equipped with OptoForce tactile sensors and a Fastrak motion tracking system. The software for evaluating different grasp quality measures is programmed in C++ and developed within the ROS middleware framework. In this thesis, three grasp quality measures have been implemented. On one side, it considers the largest-minimum resisted wrench and the volume of the grasp wrench space, where these metrics integrate limitations on finger forces, offering insights into the maximum perturbation (force and torque) each grasp can withstand in any direction. On the other side, the area of the grasp polygon defined by the contact points between the object and the fingers is also considered, taking into account the geometric relationships in the grasp and no constraints on the magnitude of the finger forces. To validate the performance of the designed system in evaluating grasp quality measures in real-world experiments, a series of tests were conducted, examining different grasp scenarios involving objects with varying geometries and materials under diverse experimental conditions. The results of the experiments are analyzed to assess the validity of the implementation. Furthermore, to enhance the system and address the challenges observed in the implementation of grasp quality measures in real grasps, potential improvements are proposed as future work.