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

Deep Learning in Vision-Based Roboti...

Yan, Mengyuan,

Linked to FindBook

Google Book

Amazon

博客來

Deep Learning in Vision-Based Robotic Manipulation: Towards Generalization and Fast Inference /

Record Type:	Electronic resources : Monograph/item
Title/Author:	Deep Learning in Vision-Based Robotic Manipulation: Towards Generalization and Fast Inference // Mengyuan Yan.
Author:	Yan, Mengyuan,
Description:	1 electronic resource (103 pages)
Notes:	Source: Dissertations Abstracts International, Volume: 83-06, Section: B.
Contained By:	Dissertations Abstracts International83-06B.
Subject:	Robots. -
Online resource:	https://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=28671147
ISBN:	9798544200000

Deep Learning in Vision-Based Robotic Manipulation: Towards Generalization and Fast Inference /
Yan, Mengyuan,

Deep Learning in Vision-Based Robotic Manipulation: Towards Generalization and Fast Inference /Mengyuan Yan. - 1 electronic resource (103 pages)

Source: Dissertations Abstracts International, Volume: 83-06, Section: B.

Robots are one of the most recent inventions in human civilizations for automation. While various robots have been used in factories to automate a wide range of tasks, they are usually programmed for specific tasks and orchestrated together with other machines and components on the manufacturing line. In the past decade, researchers are looking at bringing robots into our daily lives, and automating services such as taxis, delivery, house-works, and even medical procedures. One of the major roadblocks in making this leap is the diversity and uncertainty in the environments that the robots need to work in. Machine perception, i.e. understanding of the environment through visual, audio, and contact signals, is indispensable in such diverse and uncertain environments, and is a hard problem in itself. Further, the environment is changing, due to human activities and other factors, and robots need to react to the changes quickly. Recent developments in deep learning, especially computer vision, has brought us closer to achieving the goal of bringing robots into our daily environments. However, deep learning methods require a large amount of data with annotated labels, and new datasets and annotations need to be collected for each new task. Deep reinforcement learning algorithms have also achieved good performance on a range of locomotion or manipulation tasks, but the amount of interactions required to train most algorithms is so large that it could take days even with parallel simulation engines. Highly dataecient models and learning algorithms are needed to help robots learn faster and with less human e↵ort. Additionally, when designing a learning-based solution to a robotics task, inference speed needs to be taken into consideration so that the robot can respond to changes quickly. This thesis introduces methods to improve training data eciency and inference speed for visionbased robotic manipulation. To improve data eciency of models, we analyze properties and structures of the specific problems, and build structural biases into the models based on the insights obtained. In addition, we demonstrate self-supervised learning of the perception model on real images, enabling robots to collect their own training data without requiring human annotations. To improve robots' response speed, when learning motion policies we design learning algorithms to always explicitly learn the distribution of promising actions, instead of learning an action evaluation function which requires online optimization during runtime. The proposed methods are integrated into end-to-end systems and tested on real robots on two tasks: vision-based robotic grasping, and rope manipulation and knotting.

English

ISBN: 9798544200000Subjects--Topical Terms:

529507
Robots.

Deep Learning in Vision-Based Robotic Manipulation: Towards Generalization and Fast Inference /
LDR:04053nmm a22003733i 4500 001 2437610
005 20251124054946.5
006 m o d
007 cr|nu||||||||
008 261011s2020 miu||||||m |||||||eng d
020 $a 9798544200000
035 $a (MiAaPQD)AAI28671147
035 $a (MiAaPQD)STANFORDmd000cz6702
035 $a AAI28671147
040 $a MiAaPQD $b eng $c MiAaPQD $e rda
100 1 $a Yan, Mengyuan, $e author. $3 3828660
245 1 0 $a Deep Learning in Vision-Based Robotic Manipulation: Towards Generalization and Fast Inference / $c Mengyuan Yan.
264 1 $a Ann Arbor : $b ProQuest Dissertations & Theses, $c 2020
300 $a 1 electronic resource (103 pages)
336 $a text $b txt $2 rdacontent
337 $a computer $b c $2 rdamedia
338 $a online resource $b cr $2 rdacarrier
500 $a Source: Dissertations Abstracts International, Volume: 83-06, Section: B.
500 $a Includes supplementary digital materials.
500 $a Advisors: Bohg, Jeannette Committee members: Finn, Chelsea; Sadigh, Dorsa.
502 $b Ph.D. $c Stanford University $d 2020.
520 $a Robots are one of the most recent inventions in human civilizations for automation. While various robots have been used in factories to automate a wide range of tasks, they are usually programmed for specific tasks and orchestrated together with other machines and components on the manufacturing line. In the past decade, researchers are looking at bringing robots into our daily lives, and automating services such as taxis, delivery, house-works, and even medical procedures. One of the major roadblocks in making this leap is the diversity and uncertainty in the environments that the robots need to work in. Machine perception, i.e. understanding of the environment through visual, audio, and contact signals, is indispensable in such diverse and uncertain environments, and is a hard problem in itself. Further, the environment is changing, due to human activities and other factors, and robots need to react to the changes quickly. Recent developments in deep learning, especially computer vision, has brought us closer to achieving the goal of bringing robots into our daily environments. However, deep learning methods require a large amount of data with annotated labels, and new datasets and annotations need to be collected for each new task. Deep reinforcement learning algorithms have also achieved good performance on a range of locomotion or manipulation tasks, but the amount of interactions required to train most algorithms is so large that it could take days even with parallel simulation engines. Highly dataecient models and learning algorithms are needed to help robots learn faster and with less human e↵ort. Additionally, when designing a learning-based solution to a robotics task, inference speed needs to be taken into consideration so that the robot can respond to changes quickly. This thesis introduces methods to improve training data eciency and inference speed for visionbased robotic manipulation. To improve data eciency of models, we analyze properties and structures of the specific problems, and build structural biases into the models based on the insights obtained. In addition, we demonstrate self-supervised learning of the perception model on real images, enabling robots to collect their own training data without requiring human annotations. To improve robots' response speed, when learning motion policies we design learning algorithms to always explicitly learn the distribution of promising actions, instead of learning an action evaluation function which requires online optimization during runtime. The proposed methods are integrated into end-to-end systems and tested on real robots on two tasks: vision-based robotic grasping, and rope manipulation and knotting.
546 $a English
590 $a School code: 0212
650 4 $a Robots. $3 529507
650 4 $a Success. $3 518195
650 4 $a Curricula. $3 3422445
650 4 $a Planning. $3 552734
650 4 $a Robotics. $3 519753
690 $a 0771
710 2 $a Stanford University. $e degree granting institution. $3 3765820
720 1 $a Bohg, Jeannette $e degree supervisor.
773 0 $t Dissertations Abstracts International $g 83-06B.
790 $a 0212
791 $a Ph.D.
792 $a 2020
856 4 0 $u https://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=28671147