4.3.1 Integrated Intelligence: Robot Instruction via Interactive Grounded Learning

Abstract

People communicate using a variety of modalities, including speech, gestures, gaze, and whole body motion. To be effective, robots must be able to interpret the rich information contained in human communication. The goal of this project is to allow people to instruct robots and to teach them about objects and attributes using a combination of modalities. To achieve this goal, we are developing an interactive grounded learning system that can interpret rich human input such as speech, gesture, body motion, and gaze; and learn to identify objects based on the names and attributes people use to refer to them. This progress report summarizes the final findings and describes completed work towards a full, interactive grounded learning system.

Document Details

Document Type
DoD Grant Award
Publication Date
Jun 25, 2021
Source ID
W911NF1210197

Entities

People

  • Luke Zettlemoyer

Organizations

  • Army Contracting Command
  • United States Army
  • University of Washington

Tags

Fields of Study

  • Computer science

Readers

  • Agent-Based Social Robotics and Mobile-Assisted Learning in Virtual Environments.

Technology Areas

  • AI & ML
  • Autonomy