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 Representation Learning on Spat...

Han, Wenyu,

Linked to FindBook

Google Book

Amazon

博客來

Deep Representation Learning on Spatial and Temporal Data Towards Design and Construction Automation /

Record Type:	Electronic resources : Monograph/item
Title/Author:	Deep Representation Learning on Spatial and Temporal Data Towards Design and Construction Automation // Wenyu Han.
Author:	Han, Wenyu,
Description:	1 electronic resource (224 pages)
Notes:	Source: Dissertations Abstracts International, Volume: 85-11, Section: B.
Contained By:	Dissertations Abstracts International85-11B.
Subject:	Mechanical engineering. -
Online resource:	https://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=31240175
ISBN:	9798382757254

Deep Representation Learning on Spatial and Temporal Data Towards Design and Construction Automation /
Han, Wenyu,

Deep Representation Learning on Spatial and Temporal Data Towards Design and Construction Automation /Wenyu Han. - 1 electronic resource (224 pages)

Source: Dissertations Abstracts International, Volume: 85-11, Section: B.

Artificial intelligence-powered (AI-powered) automation span various fields such as manufacturing, construction, agriculture, medication, and design. This dissertation specifically focus on designing novel representation learning techniques on spatial and temporal data through deep neural networks over two engineering problem domains: design automation, and construction automation. The first domain is focused on automated design. The recent advancements in Artificial Intelligence Generated Content (AIGC) technologies have shown great potential in achieving automated design with lower iteration speed and higher manual intervention demand than conventional methods. While these technologies excel in rasterized formats, common in areas like art, movie editing, and webpage design, the engineering industries deal with vectorized formats, such as polygons and geometric primitives. Current AIGC methods struggle with generating vectorized content, especially in large-scale geospatial areas like city layouts, due to limited datasets, complex data statistics, and the unordered nature of vectorized data. This dissertation addresses these challenges by exploring and developing methods for automated design that can generate large-scale vectorized spatial data through deep representation learning.The second domain focuses on automated construction. Construction automation has the potential to revolutionize conventional construction methods with consistently sustained productivity, higher performance, and lower defect rates by leveraging robotics and artificial intelligence technologies. The goal of automated construction is to enable construction robots to work collaboratively with or independently of humans to build structures. This dissertation specifically addresses the challenge of autonomous manipulation of the environment by construction robots, aiming for minimal human intervention. Given the dynamic nature of construction sites, the autonomous agent must learn efficient policies under uncertainties in self-motion and environment by responding to a temporal series of sensing and actuation data. As such, this dissertation proposes a new interdependent challenge, accurate localization and strategic environment manipulation planning, particularly in scenarios where GPS and similar techniques are not available (often due to occlusions), and studies this challenge in both a simplified grid world and a realistic 3D visual simulated environment. The key contributions of this dissertation include: 1) A tree-structured autoencoder and a generative transformer based method are developed to achieve large-scale city layouts generation; 2) The interdependent challenge is addressed by decoupling representation learning from policy learning; 3) A localization-uncertainty aware deep reinforcement learning (DRL) method is designed to solve mobile object rearrangement. The introduction of these methods represents a significant stride toward the design automation of city layout generation and automated construction in highly dynamic environments.

English

ISBN: 9798382757254Subjects--Topical Terms:

649730
Mechanical engineering.
Subjects--Index Terms:

Deep representation learning

Deep Representation Learning on Spatial and Temporal Data Towards Design and Construction Automation /
LDR:04554nmm a22004213i 4500 001 2429338
005 20250430100802.5
006 m o d
007 cr|nu||||||||
008 261011s2024 miu||||||m |||||||eng d
020 $a 9798382757254
035 $a (MiAaPQD)AAI31240175
035 $a AAI31240175
040 $a MiAaPQD $b eng $c MiAaPQD $e rda
100 1 $a Han, Wenyu, $e author. $3 3817127
245 1 0 $a Deep Representation Learning on Spatial and Temporal Data Towards Design and Construction Automation / $c Wenyu Han.
264 1 $a Ann Arbor : $b ProQuest Dissertations & Theses, $c 2024
300 $a 1 electronic resource (224 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: 85-11, Section: B.
500 $a Advisors: Feng, Chen Committee members: Righetti, Ludovic; Atashzar, S. Farokh; Pinto, Lerrel.
502 $b Ph.D. $c New York University Tandon School of Engineering $d 2024.
520 $a Artificial intelligence-powered (AI-powered) automation span various fields such as manufacturing, construction, agriculture, medication, and design. This dissertation specifically focus on designing novel representation learning techniques on spatial and temporal data through deep neural networks over two engineering problem domains: design automation, and construction automation. The first domain is focused on automated design. The recent advancements in Artificial Intelligence Generated Content (AIGC) technologies have shown great potential in achieving automated design with lower iteration speed and higher manual intervention demand than conventional methods. While these technologies excel in rasterized formats, common in areas like art, movie editing, and webpage design, the engineering industries deal with vectorized formats, such as polygons and geometric primitives. Current AIGC methods struggle with generating vectorized content, especially in large-scale geospatial areas like city layouts, due to limited datasets, complex data statistics, and the unordered nature of vectorized data. This dissertation addresses these challenges by exploring and developing methods for automated design that can generate large-scale vectorized spatial data through deep representation learning.The second domain focuses on automated construction. Construction automation has the potential to revolutionize conventional construction methods with consistently sustained productivity, higher performance, and lower defect rates by leveraging robotics and artificial intelligence technologies. The goal of automated construction is to enable construction robots to work collaboratively with or independently of humans to build structures. This dissertation specifically addresses the challenge of autonomous manipulation of the environment by construction robots, aiming for minimal human intervention. Given the dynamic nature of construction sites, the autonomous agent must learn efficient policies under uncertainties in self-motion and environment by responding to a temporal series of sensing and actuation data. As such, this dissertation proposes a new interdependent challenge, accurate localization and strategic environment manipulation planning, particularly in scenarios where GPS and similar techniques are not available (often due to occlusions), and studies this challenge in both a simplified grid world and a realistic 3D visual simulated environment. The key contributions of this dissertation include: 1) A tree-structured autoencoder and a generative transformer based method are developed to achieve large-scale city layouts generation; 2) The interdependent challenge is addressed by decoupling representation learning from policy learning; 3) A localization-uncertainty aware deep reinforcement learning (DRL) method is designed to solve mobile object rearrangement. The introduction of these methods represents a significant stride toward the design automation of city layout generation and automated construction in highly dynamic environments.
546 $a English
590 $a School code: 1988
650 4 $a Mechanical engineering. $3 649730
650 4 $a Automotive engineering. $3 2181195
653 $a Deep representation learning
653 $a Complex data statistics
653 $a Conventional methods
653 $a Human intervention
690 $a 0548
690 $a 0800
690 $a 0540
710 2 $a New York University Tandon School of Engineering. $b Mechanical & Aerospace Engineering. $e degree granting institution. $3 3817056
720 1 $a Feng, Chen $e degree supervisor.
773 0 $t Dissertations Abstracts International $g 85-11B.
790 $a 1988
791 $a Ph.D.
792 $a 2024
856 4 0 $u https://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=31240175