An Architecture for Shared Autonomy via Optimal Control
Abstract
Shared Autonomy studies how humans and robots can work together to solve problems together. We focus on the problem of a human teleoperator controlling a robot to perform a complex manipulation task. Traditional paradigms focus on transferring the human s inputs directly to the robot. Although effective, this results in operator fatigue, and does not enable the system to learn from the human s motions. We propose a paradigm where the system learns via apprenticeship how to perform the task as it is being controlled by the human, gradually taking over more aspects of the task as it gets more proficient and confident.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Oct 01, 2019
- Source ID
- W911NF1510312
Entities
People
- Siddhartha Srinivasa
Organizations
- Army Contracting Command
- Defense Advanced Research Projects Agency
- Massachusetts Institute of Technology