Human-Guided Robot Manipulation
Robots must be able to manipulate objects and physically interact with their surrounding environment, however robust, dexterous manipulation remains an open challenge in robotics research. Robots can autonomously learn to interact with objects on their own, but their ability to manipulate objects with complex or changing dynamics is limited.
Some approaches to achieve human-like dexterity include designing highly complex robotic hands and mapping human forces and motion onto robotic hands. While these approaches have played an integral role in advancing robot manipulation, there is still a major limitation that is often overlooked, which is that the mechanics of a human and robot hand are fundamentally different. The interactive dynamics between a human hand and an object differ from those between a robot hand and the same object. Without modification to the human motion being mapped, the gripper will not be able to perform the same task due to the differences in its interaction mechanics.
The goal of this project is to improve how robots learn how to manipulate objects and perform complex interaction tasks from human guidance by (1) accounting for differences between human and robot mechanics and (2) developing novel methods and hardware that make human guidance more intuitive.