Start Date

19-5-2023 3:40 PM

End Date

19-5-2023 3:50 PM

Document Type

Extended Abstract

Keywords

Grasping; Dexterous Manipulation; In-Hand Manipulation

Description

Grasping multiple objects at once from a pile is common for humans. It makes us efficient in pick and transfer tasks. It is essential for a robot to gain multi-object grasping capability (MOG). This paper defines the multi-object grasping problem and introduces several novel multi-object grasping techniques. These techniques include probability-based pre-grasp potential calculation, a stochastic flexing/extending routine, obtaining end-grasp types, and estimating the number of objects in a grasp. It also proposes a new stochastic grasping strategy for grasping a desired number of objects.

DOI

https://doi.org/10.5038/AWXI6560

Download

Included in

Robotics Commons

Share

COinS
 
May 19th, 3:40 PM May 19th, 3:50 PM

Multi-Object Grasping -- Stochastic Grasping from a Pile

Grasping multiple objects at once from a pile is common for humans. It makes us efficient in pick and transfer tasks. It is essential for a robot to gain multi-object grasping capability (MOG). This paper defines the multi-object grasping problem and introduces several novel multi-object grasping techniques. These techniques include probability-based pre-grasp potential calculation, a stochastic flexing/extending routine, obtaining end-grasp types, and estimating the number of objects in a grasp. It also proposes a new stochastic grasping strategy for grasping a desired number of objects.