東華大學圖書館 |

Language: English

Help

回圖書館首頁

手機版館藏查詢

Back

Switch To: Labeled | MARC Mode | ISBD

Learning by imitation and exploratio...

University of Washington.

Linked to FindBook

Google Book

Amazon

博客來

Learning by imitation and exploration: Bayesian models and applications in humanoid robotics.

Record Type:	Language materials, printed : Monograph/item
Title/Author:	Learning by imitation and exploration: Bayesian models and applications in humanoid robotics./
Author:	Grimes, David B.
Description:	109 p.
Notes:	Adviser: Rajesh P. N. Rao.
Contained By:	Dissertation Abstracts International68-12B.
Subject:	Computer Science. -
Online resource:	http://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=3293481
ISBN:	9780549377320

Learning by imitation and exploration: Bayesian models and applications in humanoid robotics.
Grimes, David B.

Learning by imitation and exploration: Bayesian models and applications in humanoid robotics. - 109 p.

Adviser: Rajesh P. N. Rao.

Thesis (Ph.D.)--University of Washington, 2007.

Ultimately, this thesis seeks to contribute novel ideas which one day may form the basis for a powerful human-robot interface which allows people to quickly and effortlessly train robots to perform new skills.

ISBN: 9780549377320Subjects--Topical Terms:

626642
Computer Science.

Learning by imitation and exploration: Bayesian models and applications in humanoid robotics.
LDR:03369nam 2200301 a 45 001 856726
005 20100709
008 100709s2007 ||||||||||||||||| ||eng d
020 $a 9780549377320
035 $a (UMI)AAI3293481
035 $a AAI3293481
040 $a UMI $c UMI
100 1 $a Grimes, David B. $3 1023580
245 1 0 $a Learning by imitation and exploration: Bayesian models and applications in humanoid robotics.
300 $a 109 p.
500 $a Adviser: Rajesh P. N. Rao.
500 $a Source: Dissertation Abstracts International, Volume: 68-12, Section: B, page: 8128.
502 $a Thesis (Ph.D.)--University of Washington, 2007.
520 $a Ultimately, this thesis seeks to contribute novel ideas which one day may form the basis for a powerful human-robot interface which allows people to quickly and effortlessly train robots to perform new skills.
520 $a Learning by imitation is an important mechanism for rapid acquisition of new skills in humans and robots. A critical requirement for learning by imitation is the ability to reason under uncertainty. Uncertainty arises during the process of observing the teacher as well as from the imitator's own dynamics and interactions with the environment. This dissertation introduces new probabilistic methods for learning novel, generalizable behaviors in a humanoid robot via imitation. At the heart of my approach is a proposed algorithm for selecting actions based on probabilistic inference in a learned Bayesian network. This inference-based action selection technique affords an efficient, straightforward method of exploiting rich, yet uncertain, sensory data gathered by the robot.
520 $a Many existing methods for planning robotic actions require an engineer to explicitly model the complex physics of the robot and its environment. This process can be costly, tedious, error-prone, brittle, and inflexible to changes in the environment or the robot. The method I propose involves learning a predictive model of the robot's dynamics, represented directly in terms of sensor measurements, solely from exploration. Experiments are performed with a Fujitsu HOAP-2 25-degrees-of-freedom humanoid robot and the Webots dynamic simulation software. I present results demonstrating that the robot can learn dynamically stable, full-body imitative motions simply by observing a human demonstrator and performing explorative learning. Additional results show how the inference-based action selection technique can be used for policy learning, where sensory feedback can be used to adapt behavior online. I present policy learning results for a lifting behavior (learned via imitation) that generalizes to a wide range of objects of novel, unknown density. Besides imitation-based learning, this dissertation makes other contributions to the emerging area of robotic learning. First, intractability due to very high-dimensional state and control spaces is tackled using dimensionality reduction techniques. Second, nonparametric techniques are introduced to handle the problem of learning and inference with continuous-valued random variables.
590 $a School code: 0250.
650 4 $a Computer Science. $3 626642
650 4 $a Engineering, Robotics. $3 1018454
690 $a 0771
690 $a 0984
710 2 $a University of Washington. $3 545923
773 0 $t Dissertation Abstracts International $g 68-12B.
790 $a 0250
790 1 0 $a Rao, Rajesh P. N., $e advisor
791 $a Ph.D.
792 $a 2007
856 4 0 $u http://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=3293481