Probabilistic Approach to Sensor-based Grasping

In this paper, we present a probabilistic framework for grasping. In the framework, we consider grasp and object attributes, on-line sensor information and the stability of a grasp, through probabilistic models. We describe how sensorbased grasp planning can be formulated in a probabilistic framework and how information about object attributes can be updated simultaneously using on-line sensor information gained during grasping. The feasibility and advantages of the framework are demonstrated in a 2D simulation environment, with simulated tactile sensors used to update object information. In the demonstration, particle filters are used to model the evolving probability distributions.