Abstract:
As robotic systems become increasingly capable and commercially available, particularly in the form of humanoids and dexterous hands, enabling them to physically interact with their environment remains a fundamental challenge. Humans effortlessly use contact with their surroundings to perform agile movements and manipulate both delicate and heavy objects. For robots, in contrast, effectively leveraging physical contact, whether through full-body interactions for enhanced agility, or towards contact-rich manipulation with robot hands, is still a complex and largely unsolved problem.

This thesis addresses key challenges in learning and control for physical interaction for agile robots and dexterous manipulation. Specifically, it investigates how to effectively capture human demonstrations of contact-rich and dynamic tasks and how to translate them into robot capabilities. Analysis reveals that relatively simple models can sufficiently represent human dynamic interactions at an abstract level (e.g., hand motion and contact forces relative to the center of mass). Combined with reflex-like controllers, these simple models can be used to recreate dynamic physical interaction behaviors in robots.

Building on these reflex-based control modules for whole-body interactions, this thesis aims to extend the same principles to dexterous robot hands. Here, similar abstract representations can be defined, for example, fingertip positions or contact points relative to the wrist. Just as the human body exhibits compliant behavior during physical interactions, the hand and fingers can also be treated as a compliant system. This view enables the development of comparable control strategies for manipulation. Adding even more compliance, the use of soft materials in robot hands offers the potential for inherent safety, adaptability, and robustness through the natural flexibility these materials provide.

Another key objective of this work is to identify and integrate the necessary sensing modalities, such as proprioception and tactile feedback, into soft robotic hand hardware. The proposed work aims to characterize and benchmark multimodal sensor data, evaluate its utility for manipulation, and develop new methods to incorporate this feedback into both learning-based and model-based control frameworks.

Thesis Committee Members:
Nancy Pollard (Chair)
Oliver Kroemer
Zackory Erickson
Joohyung Kim (University of Illinois Urbana-Champaign)

More Information