Learning to Select Primitives and Generate Sub-goals from Practice

Darrin C. Bentivegna; Christopher G. Atkeson; Gordon Cheng

Learning to Select Primitives and Generate Sub-goals from Practice

Darrin C. Bentivegna, Christopher G. Atkeson, Gordon Cheng

Research output: Contribution to conference › Paper › peer-review

4 Scopus citations

Abstract

This paper focuses on learning to select behavioral primitives and generate sub-goals from practicing a task. We present a novel algorithm that combines Q-learning and a locally weighted learning method to improve primitive selection and sub-goal generation. We demonstrate this approach applied to the tilt maze task (see video). Our robot initially learns to perform this task using learning from observation, and then learns from practice.

Original language	English
Pages	946-953
Number of pages	8
State	Published - 2003
Externally published	Yes
Event	2003 IEEE/RSJ International Conference on Intelligent Robots and Systems - Las Vegas, NV, United States Duration: 27 Oct 2003 → 31 Oct 2003

Conference

Conference	2003 IEEE/RSJ International Conference on Intelligent Robots and Systems
Country/Territory	United States
City	Las Vegas, NV
Period	27/10/03 → 31/10/03

Cite this

@conference{c06f9a65406143cfb1520b9c1d30bfe1,

title = "Learning to Select Primitives and Generate Sub-goals from Practice",

abstract = "This paper focuses on learning to select behavioral primitives and generate sub-goals from practicing a task. We present a novel algorithm that combines Q-learning and a locally weighted learning method to improve primitive selection and sub-goal generation. We demonstrate this approach applied to the tilt maze task (see video). Our robot initially learns to perform this task using learning from observation, and then learns from practice.",

author = "Bentivegna, {Darrin C.} and Atkeson, {Christopher G.} and Gordon Cheng",

year = "2003",

language = "English",

pages = "946--953",

note = "2003 IEEE/RSJ International Conference on Intelligent Robots and Systems ; Conference date: 27-10-2003 Through 31-10-2003",

}

TY - CONF

T1 - Learning to Select Primitives and Generate Sub-goals from Practice

AU - Bentivegna, Darrin C.

AU - Atkeson, Christopher G.

AU - Cheng, Gordon

PY - 2003

Y1 - 2003

N2 - This paper focuses on learning to select behavioral primitives and generate sub-goals from practicing a task. We present a novel algorithm that combines Q-learning and a locally weighted learning method to improve primitive selection and sub-goal generation. We demonstrate this approach applied to the tilt maze task (see video). Our robot initially learns to perform this task using learning from observation, and then learns from practice.

AB - This paper focuses on learning to select behavioral primitives and generate sub-goals from practicing a task. We present a novel algorithm that combines Q-learning and a locally weighted learning method to improve primitive selection and sub-goal generation. We demonstrate this approach applied to the tilt maze task (see video). Our robot initially learns to perform this task using learning from observation, and then learns from practice.

UR - http://www.scopus.com/inward/record.url?scp=0348040468&partnerID=8YFLogxK

M3 - Paper

AN - SCOPUS:0348040468

SP - 946

EP - 953

T2 - 2003 IEEE/RSJ International Conference on Intelligent Robots and Systems

Y2 - 27 October 2003 through 31 October 2003

ER -

Learning to Select Primitives and Generate Sub-goals from Practice

Abstract

Conference

Other files and links

Fingerprint

Cite this