Mobile Manipulation Platform for Language-Based Shared Autonomy
Abstract
We propose to develop an outdoor-capable robotic mobile manipulator platform to support ongoing and future DoD research programs at the Toyota Technological Institute at Chicago (TTIC). The proposed research will address challenges that are fundamental to the DoDÕs focus on soldier-robot teaming and human-machine collaboration more generally. This proposal will enable the development and realization of models and algorithms that enable a soldier to remotely command a robotic mobile manipulator to perform dexterous handling tasks in a shared autonomy framework. Integral and unique to our proposed effort is the use of natural language utterances as a flexible, efficient ÒsensorÓ stream as well as a command and control medium. During an initial training phase, the soldier can use natural language as a supervisory signal to teach the robot how to carry out complex manipulation tasks. Having learned these new skills, the soldier can subsequently command the robot to perform the task simply by speaking a natural language instruction. Successful realization of our proposed shared autonomy framework raises fundamental questions related to grounded language acquisition, machine perception, and human-robot interaction. Ongoing and proposed research address these challenges through three primary thrusts in natural language understanding, mapping, and language-guided manipulation under DoD efforts lead by PI Walter at TTIC. The proposed platform consisting of a Clearpath Husky Unmanned Ground Vehicle equipped with a Universal Robots UR5 arm, Robotiq 3-Finger Gripper, 3D LIDAR, and stereo camera, will be integral to these efforts.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Sep 11, 2018
- Source ID
- W911NF1710188
Entities
People
- Matthew J Walter
Organizations
- Army Contracting Command
- Toyota Technological Institute at Chicago
- United States Army