서브메뉴
검색
Perception and Controls for Task-Flexible Manufacturing Robots
Perception and Controls for Task-Flexible Manufacturing Robots
Detailed Information
- 자료유형
- 학위논문 서양
- 최종처리일시
- 20260202105228
- ISBN
- 9798291566886
- DDC
- 621
- 서명/저자
- Perception and Controls for Task-Flexible Manufacturing Robots
- 발행사항
- [Sl] : University of Michigan, 2025
- 발행사항
- Ann Arbor : ProQuest Dissertations & Theses, 2025
- 형태사항
- 104 p
- 주기사항
- Source: Dissertations Abstracts International, Volume: 87-03, Section: B.
- 주기사항
- Advisor: Fazeli, Nima;Shih, Albert.
- 학위논문주기
- Thesis (Ph.D.)--University of Michigan, 2025.
- 초록/해제
- 요약In practice, automation in industrial manufacturing succeeds with limited task flexibility. Hardware is designed with a few specific purposes in mind. This has costly side-effects when designs for products are changed or manufacturing methods are improved. The rapidly evolving field of robotics opens up new possibilities for systems that are task-flexible. In this dissertation, I develop and demonstrate intelligent, task-flexible systems for manufacturing and robotic manipulation. These systems rely heavily on perception and controls methods at the intersection of recently introduced machine learning-based methods, and traditional computer vision, geometrical, and model-based methods. While the chapters presented in this thesis are diverse, they all illustrate ways to improve performance, certainty, and precision in industrial robotics systems where these aspects are vital. I first propose a framework for for compensation and control in robotic high-viscosity fluid deposition, a process tied to additive manufacturing and sealant/adhesive dispensing. This method relies on a learned but generalizable model, alongside classical model predictive control. A novel on-hand vision-based flow rate sensor is introduced as a perception component to solving this problem, and this sensor relies on learned segmentation models. The following chapters rely on vision-based tactile sensing for their perception component. Tactile sensing provides much flexibility in comparison to visual sensing, as it is subject to less of a distribution shift and resistant to heavily occluded environments. I next propose a framework for transferring compliant robot behavior, typical for human-robot collaboration, across embodiments. Through this framework, impedance control can be replicated on robots with just tactile sensing and no force-torque sensing. Finally, I consult the problem of precise robotic assembly, such as threading or low-clearance non-chamfered insertion. Modern robotic methods like behavior cloning have previously demonstrated low success rates for precise tasks using purely tactile sensing. This is improved upon using Grasped Object Manifold Projection (GOMP), proposed in the final work of this thesis. GOMP demonstrates the strength of geometric methods applied on top of state-of-the-art imitation learning. In summary, this dissertation proposes generalizable, task-flexible methods in manufacturing and manipulation, combining the advantages of machine learning with classical engineering.
- 일반주제명
- Mechanical engineering
- 일반주제명
- Robotics
- 일반주제명
- Engineering
- 키워드
- Tactile sensing
- 키워드
- Behavior cloning
- 기타저자
- University of Michigan Mechanical Engineering
- 기본자료저록
- Dissertations Abstracts International. 87-03B.
- 전자적 위치 및 접속
- 로그인 후 원문을 볼 수 있습니다.
MARC
008260126s2025 us c eng d■001000017359870
■00520260202105228
■006m o d
■007cr#unu||||||||
■020 ▼a9798291566886
■035 ▼a(MiAaPQ)AAI32271867
■035 ▼a(MiAaPQ)umichrackham006509
■040 ▼aMiAaPQ▼cMiAaPQ
■0820 ▼a621
■1001 ▼avan den Bogert, William.
■24510▼aPerception and Controls for Task-Flexible Manufacturing Robots
■260 ▼a[Sl]▼bUniversity of Michigan▼c2025
■260 1▼aAnn Arbor▼bProQuest Dissertations & Theses▼c2025
■300 ▼a104 p
■500 ▼aSource: Dissertations Abstracts International, Volume: 87-03, Section: B.
■500 ▼aAdvisor: Fazeli, Nima;Shih, Albert.
■5021 ▼aThesis (Ph.D.)--University of Michigan, 2025.
■520 ▼aIn practice, automation in industrial manufacturing succeeds with limited task flexibility. Hardware is designed with a few specific purposes in mind. This has costly side-effects when designs for products are changed or manufacturing methods are improved. The rapidly evolving field of robotics opens up new possibilities for systems that are task-flexible. In this dissertation, I develop and demonstrate intelligent, task-flexible systems for manufacturing and robotic manipulation. These systems rely heavily on perception and controls methods at the intersection of recently introduced machine learning-based methods, and traditional computer vision, geometrical, and model-based methods. While the chapters presented in this thesis are diverse, they all illustrate ways to improve performance, certainty, and precision in industrial robotics systems where these aspects are vital. I first propose a framework for for compensation and control in robotic high-viscosity fluid deposition, a process tied to additive manufacturing and sealant/adhesive dispensing. This method relies on a learned but generalizable model, alongside classical model predictive control. A novel on-hand vision-based flow rate sensor is introduced as a perception component to solving this problem, and this sensor relies on learned segmentation models. The following chapters rely on vision-based tactile sensing for their perception component. Tactile sensing provides much flexibility in comparison to visual sensing, as it is subject to less of a distribution shift and resistant to heavily occluded environments. I next propose a framework for transferring compliant robot behavior, typical for human-robot collaboration, across embodiments. Through this framework, impedance control can be replicated on robots with just tactile sensing and no force-torque sensing. Finally, I consult the problem of precise robotic assembly, such as threading or low-clearance non-chamfered insertion. Modern robotic methods like behavior cloning have previously demonstrated low success rates for precise tasks using purely tactile sensing. This is improved upon using Grasped Object Manifold Projection (GOMP), proposed in the final work of this thesis. GOMP demonstrates the strength of geometric methods applied on top of state-of-the-art imitation learning. In summary, this dissertation proposes generalizable, task-flexible methods in manufacturing and manipulation, combining the advantages of machine learning with classical engineering.
■590 ▼aSchool code: 0127.
■650 4▼aMechanical engineering
■650 4▼aRobotics
■650 4▼aEngineering
■653 ▼aManufacturing automation
■653 ▼aAdditive manufacturing
■653 ▼aTactile sensing
■653 ▼aBehavior cloning
■653 ▼aPerception and controls
■690 ▼a0771
■690 ▼a0548
■690 ▼a0800
■690 ▼a0537
■71020▼aUniversity of Michigan▼bMechanical Engineering.
■7730 ▼tDissertations Abstracts International▼g87-03B.
■790 ▼a0127
■791 ▼aPh.D.
■792 ▼a2025
■793 ▼aEnglish
■85640▼uhttp://www.riss.kr/pdu/ddodLink.do?id=T17359870▼nKERIS▼z이 자료의 원문은 한국교육학술정보원에서 제공합니다.
Preview
Export
ChatGPT Discussion
AI Recommended Related Books
Подробнее информация.
- Бронирование
- не существует
- моя папка
- Первый запрос зрения
- Non-Book Loan Application
- Nighttime Book Loan Application
Available after logging in.


