Colloquium Details
Confidence-Based Robot Policy Learning from Demonstration
Author: | Sonia Chernova Carnegie Mellon University |
---|---|
Date: | April 09, 2009 |
Time: | 15:30 |
Location: | 220 Deschutes |
Host: | Jun Li |
Abstract
The problem of learning a policy, a task representation mapping from world states to actions, lies at the heart of many robotics applications. One approach to acquiring a task policy is learning from demonstration, an interactive technique in which a robot learns a policy based on example state to action mappings provided by a human teacher.
In this talk, I will introduce Confidence-Based Autonomy, a mixed-initiative single robot demonstration learning algorithm that enables the robot and teacher to jointly control the learning process and selection of demonstration training data. Our algorithm enables the robot to identify the need for and request demonstrations for specific parts of the state space based on confidence thresholds characterizing the uncertainty of the learned policy. The robot's demonstration requests are complemented by the teacher's ability to provide supplementary corrective demonstrations in error cases.
Based on this single-robot algorithm, I will present a task and platform independent multi-robot demonstration learning framework that enables a single person to teach multiple robots to perform collaborative tasks. I will introduce two methods of teaching communication-based coordination, through the use of active communication actions and passive state sharing, and demonstrate the scalability of this approach to tasks involving up to seven robots. I will conclude the talk with a discussion of the potential that learning from demonstration research has for making robots more accessible to everyday people.
Biography
Sonia Chernova is a doctoral student in the Computer Science Department at Carnegie Mellon University, working under Manuela Veloso. She received her computer science B.S. with honors from CMU in 2003. Her research interests focus on the development of algorithms that enable robots to interact with and assist people in complex, dynamic environments. In the development of such algorithms, her interests span artificial intelligence, reasoning under uncertainty, learning, multi-robot systems, and human-robot interaction.