Marianna Madry

Publications

From Object Categories to Grasp Transfer Using Probabilistic Reasoning

Marianna Madry, Dan Song and Danica Kragic.
In IEEE International Conference on Robotics and Automation (ICRA 2012), May 2012. To appear

Introduction

The main objective of this work is to enable transfer of grasp knowledge between object categories, defined using both their physical properties and functionality. It is a challenging problem due to the fact that a number of objects having similar physical properties afford different tasks. An example can be a screwdriver and a carrot that are structurally alike, but only the former can be used as a tool, or a ball and an orange where only the former affords playing.

       

Moreover, not just any grasp can be applied on the object. The action space is constrained by the task. When grasping a cup to pour a liquid a robot hand should not block the opening. Similarly, when using a screwdriver as a tool the robot hand should grasp the handle.

The system is built upon an:

• active scene segmentation module able to generate object hypotheses and segment them from the background in real-time,
• object categorization system using integration of 2D and 3D cues, and
• probabilistic grasp reasoning system.

Individual object hypotheses are first generated, categorized and then used as the input to a grasp generation and transfer system that encodes task, object and action properties. The experimental evaluation compares individual 2D and 3D categorization approaches with the integrated system, and it demonstrates the usefulness of the categorization in task-based grasping and grasp transfer.

Videos

We present experimental results of generating the task-constrained grasping points for the real scene. Click on one of the blue images below to watch a video for a corresponding scene and task.