Value of Assistance for Grasping

Mohammad Masarwy, Yuval Goshen, David Dovrat, Sarah Keren

Keywords: Grasping, Active Sensing, Manipulation, Helpfulness, Assistance, Value of Assitance

Abstract: In many realistic settings, a robot is tasked with grasping an object without knowing its exact pose and needs to rely instead on a probabilistic estimation of the pose in order to decide how to attempt the grasp. We support settings in which it is possible to provide the robot with an observation of the object before a grasp is attempted. However, since this possibility is limited, there is a need to predict the expected benefit of possible sensing actions before choosing which one to perform. For this, we offer new ways to compute Value of Assistance (VOA) for grasping to predict the expected effect a sensing action will have on the probability of a successful grasp. Our empirical evaluation shows how our suggested VOA measure can be used to identify helpful sensing actions.

Submission Number: 2