RI | Publications | Identifying Trajectory Classes in Dynamic Tasks

Text only version of this site

Identifying Trajectory Classes in Dynamic Tasks
S. Anderson and S. Srinivasa
International Symposium on Approximate Dynamic Programming and Reinforcement Learning, April, 2007.

Jump to: Download | Abstract | Text Reference | BibTeX Reference

Download [Help]

This publication is available, but is not online for the following reason: IEEE Holds Copyright.
To request a copy, please contact (riweb-pubs@ri.cmu.edu).

Abstract

Using domain knowledge to decompose difficult control problems is a widely used technique in robotics. Previous work has automated the process of identifying some qualitative behaviors of a system, finding a decomposition of the system based on that behavior, and constructing a control policy based on that decomposition. We introduce a novel method for automatically finding decompositions of a task based on observing the behavior of a preexisting controller. Unlike previous work, these decompositions define reparameterizations of the state space that can permit simplified control of the system.

Text Reference

S. Anderson and S. Srinivasa, "Identifying Trajectory Classes in Dynamic Tasks," International Symposium on Approximate Dynamic Programming and Reinforcement Learning, April, 2007.

BibTeX Reference

@inproceedings{Anderson_2007_5606,
   author = "Stuart Anderson and Siddhartha Srinivasa",
   title = "Identifying Trajectory Classes in Dynamic Tasks",
   booktitle = "International Symposium on Approximate Dynamic Programming and Reinforcement Learning",
   month = "April",
   year = "2007"
}