Learning Task Skills and Goals Simultaneously from Physical Interaction

TL;DR

This paper presents a framework for robots to learn manipulation skills and long-term goals simultaneously through physical interaction, enhancing their ability to align with human objectives in complex tasks.

Contribution

It introduces a novel formulation that enables concurrent learning of task skills and goals from physical human-robot interactions, addressing previous limitations.

Findings

Framework successfully infers long-term goals from interactions

Robots adapt behaviors to align with human objectives

Demonstrates feasibility in complex manipulation tasks

Abstract

In real-world human-robot systems, it is essential for a robot to comprehend human objectives and respond accordingly while performing an extended series of motor actions. Although human objective alignment has recently emerged as a promising paradigm in the realm of physical human-robot interaction, its application is typically confined to generating simple motions due to inherent theoretical limitations. In this work, our goal is to develop a general formulation to learn manipulation functional modules and long-term task goals simultaneously from physical human-robot interaction. We show the feasibility of our framework in enabling robots to align their behaviors with the long-term task objectives inferred from human interactions.