Sensors and methods for the evaluation of grasping

Abstract

Evaluation of robot grasping consists in assessing whether a grasp on a target object meets certain desirable contact properties related with its robustness, reliability, and stability. Traditionally, most works on robot grasp planning have focused on pre-grasp analysis, where the target objects and tasks are considered in order to provide a set of feasible candidate grasps. Evaluation has been restricted to rank them in order to ease the selection. Form and force closure criteria have been defined to determine the grasp feasibility and several quality metrics have been developed to evaluate the goodness of a given grasp. However, grasp evaluation is not only concerned with the planning stage. It is also relevant while the grasping action is occurring or has been completed. For this kind of approaches a variety of sensor modalities can be used in order to assess and improve the state of the grasp. The most common sensors are contact based, though visual input and proximity sensors are also used. The aim of sensing capabilities in this stage is to provide feedback in order to improve or adapt the robot hand to the objects and environment conditions. In addition sensing information can also be used to provide post-grasp quality tests.This chapter gives an overview of sensor-based grasp evaluation. On the first place a survey on the major sensor modalities and technologies is provided, with special focus on contact-based sensors. Commercially available sensors and their applications are described. This survey is completed with a number of illustrative cases in which sensor feedback is used to evaluate the goodness of grasping and to improve its execution.