Leveraging Deformations in Soft Objects and Soft Robots for Dexterity

Uksang Yoo

Keywords: Deformable Object Manipulation, Soft Robots, Dexterous Manipulation

TL;DR: Dexterous manipulation becomes easier when robots reason about deformation explicitly and not avoid it.

Abstract: In this talk, I will present a perspective on dexterous manipulation centered on explicitly reasoning about deformation. I will begin by arguing that many real-world manipulation tasks—ranging from cables and cloth to clay and hair—require not avoiding deformation, but deliberately inducing and exploiting it. This motivates treating deformation as a structured, task-relevant state rather than a by-product of interaction. I will then discuss a series of works that develop this idea across domains. First, I will show how mechanics-inspired representations enable robust perception and control of deformable objects, allowing robots to manipulate materials such as clay and hair under uncertainty. Next, I will present how similar principles apply to soft robots, where estimating the robot’s own deformation enables reliable sensing and control. I will then highlight how deformation can serve as an intermediate representation for skill acquisition, enabling more effective learning from demonstration. Finally, I will conclude with recent directions that move beyond deformation to explicitly model contact forces from human demonstrations, aiming to learn transferable interaction representations that generalize across objects, tasks, and embodiments.

Email Sharing: We authorize the sharing of all author emails with Program Chairs.

Data Release: We authorize the release of our submission and author names to the public in the event of acceptance.

Submission Number: 16