Computer Science Department , MS Thesis Presentation, Isabella Feeney "Visual Contact Detection for In-Hand Manipulation"

Isabella Feeney

MS student

WPI – Computer Science Department 

Monday, August 11, 2025

Time: 12:00 pm – 1:00 PM

Zoom Link: https://wpi.zoom.us/j/99058582614 

Advisor :  Prof. Berk Calli, WPI – RBE Department

Reader : Prof. Joshua Cuneo , WPI – Computer Science Department

Abstract:

In-hand manipulation involves changing the pose of an object within a hand without

the need to put it down and regrasp it. While this is a common skill for humans,

such as picking up a cell phone and shifting its position to view the screen, in-

hand manipulation remains a significant challenge for robots. This project proposes

using visual pose-tracking tools to accurately estimate the position of the object

in hand, enabling closed-loop control of in-hand manipulation for robotic systems.

We develop an algorithm to account for the physical constraints present during

in-hand manipulation, increasing the accuracy of the estimation. Specifically, we

account for the knowledge that the surface of the object must be in contact with the

robotic hand, but not intersecting it, while the object is grasped. We then conduct

several experiments to determine optimal heuristics for our estimation algorithm,

prioritizing solutions that are closest to the original predicted pose of the object.

Using this baseline, we move on to testing our method on manipulation tasks using a

real robotic gripper. Finally, we evaluate the performance of this algorithm against

the standard pose estimation method.