서브메뉴
검색
Contact-Aware Learning in Robotic Grasping, Manipulation and Sensing
Contact-Aware Learning in Robotic Grasping, Manipulation and Sensing
상세정보
- 자료유형
- 학위논문 서양
- 최종처리일시
- 20250211151000
- ISBN
- 9798384451952
- DDC
- 629.8
- 저자명
- Zhu, Xinghao.
- 서명/저자
- Contact-Aware Learning in Robotic Grasping, Manipulation and Sensing
- 발행사항
- [Sl] : University of California, Berkeley, 2024
- 발행사항
- Ann Arbor : ProQuest Dissertations & Theses, 2024
- 형태사항
- 138 p
- 주기사항
- Source: Dissertations Abstracts International, Volume: 86-03, Section: B.
- 주기사항
- Advisor: Tomizuka, Masayoshi.
- 학위논문주기
- Thesis (Ph.D.)--University of California, Berkeley, 2024.
- 초록/해제
- 요약In an era where robotic manipulators are increasingly sought after for customized production and household services, this dissertation delves into the development of advanced skills and dexterity in robot manipulations. It proposes an innovative approach by integrating physical principles, particularly contact mechanics, into robot learning. This integration aims to enhance data efficiency and reliability, moving beyond the current paradigm where learning agents heavily depend on vast data and extensive exploration.The dissertation unfolds in three aspects of contact-aware robot learning. Firstly, it pioneers the development of robust and sample-efficient grasping frameworks. In Chapter 2, a contrastive grasp planning module is introduced to mitigate the effects of camera noise and simulation-to-reality gap, thereby improving grasp robustness. Chapter 3 presents the Maximum Likelihood Grasp Sampling Loss, achieving an eightfold reduction in training sample requirements compared to existing methods. Additionally, Chapter 4 explores grasping planning for multi-fingered hands, expanding the versatility of robotic manipulators. The second aspect delves into the integration of contact planning into a spectrum of manipulation tasks. Chapter 5 proposes contact-aware learning from demonstrations. This approach allows robots to assimilate skills by observing human demonstrations, effectively accelerating robotic skill acquisition. Chapter 6 explores the concept of safe contact in robotic operations, investigating strategies that permit contact with obstacles while ensuring safety. In Chapter 7, an intelligent robotic assembly framework is introduced, featuring multi-level reasoning that combines sequence reasoning transformers and meticulous planning of contact points. The third aspect focuses on the sensing of contact through vision-based tactile sensors, as discussed in Chapter 8. This chapter presents a method for reconstructing contact profiles from image imprints captured by these sensors, providing a crucial feedback mechanism during manipulation tasks.Collectively, these contributions present a suite of novel grasping, manipulation, and assembly strategies. They are designed to reduce reliance on hand-engineering, thereby improving the efficiency and stability of robotic systems in diverse applications. This comprehensive body of work demonstrates the feasibility and benefits of integrating contact into robot learning. The effectiveness and practical applicability of these methods are rigorously validated through a series of simulations and real-world experiments involving various manipulators and hands.
- 일반주제명
- Robotics
- 일반주제명
- Computer science
- 일반주제명
- Mechanical engineering
- 일반주제명
- Computer engineering
- 키워드
- Machine learning
- 키워드
- Robotic grasping
- 키워드
- Sensors
- 기타저자
- University of California, Berkeley Mechanical Engineering
- 기본자료저록
- Dissertations Abstracts International. 86-03B.
- 전자적 위치 및 접속
- 로그인 후 원문을 볼 수 있습니다.
MARC
008250123s2024 us c eng d■001000017160341
■00520250211151000
■006m o d
■007cr#unu||||||||
■020 ▼a9798384451952
■035 ▼a(MiAaPQ)AAI30993899
■040 ▼aMiAaPQ▼cMiAaPQ
■0820 ▼a629.8
■1001 ▼aZhu, Xinghao.
■24510▼aContact-Aware Learning in Robotic Grasping, Manipulation and Sensing
■260 ▼a[Sl]▼bUniversity of California, Berkeley▼c2024
■260 1▼aAnn Arbor▼bProQuest Dissertations & Theses▼c2024
■300 ▼a138 p
■500 ▼aSource: Dissertations Abstracts International, Volume: 86-03, Section: B.
■500 ▼aAdvisor: Tomizuka, Masayoshi.
■5021 ▼aThesis (Ph.D.)--University of California, Berkeley, 2024.
■520 ▼aIn an era where robotic manipulators are increasingly sought after for customized production and household services, this dissertation delves into the development of advanced skills and dexterity in robot manipulations. It proposes an innovative approach by integrating physical principles, particularly contact mechanics, into robot learning. This integration aims to enhance data efficiency and reliability, moving beyond the current paradigm where learning agents heavily depend on vast data and extensive exploration.The dissertation unfolds in three aspects of contact-aware robot learning. Firstly, it pioneers the development of robust and sample-efficient grasping frameworks. In Chapter 2, a contrastive grasp planning module is introduced to mitigate the effects of camera noise and simulation-to-reality gap, thereby improving grasp robustness. Chapter 3 presents the Maximum Likelihood Grasp Sampling Loss, achieving an eightfold reduction in training sample requirements compared to existing methods. Additionally, Chapter 4 explores grasping planning for multi-fingered hands, expanding the versatility of robotic manipulators. The second aspect delves into the integration of contact planning into a spectrum of manipulation tasks. Chapter 5 proposes contact-aware learning from demonstrations. This approach allows robots to assimilate skills by observing human demonstrations, effectively accelerating robotic skill acquisition. Chapter 6 explores the concept of safe contact in robotic operations, investigating strategies that permit contact with obstacles while ensuring safety. In Chapter 7, an intelligent robotic assembly framework is introduced, featuring multi-level reasoning that combines sequence reasoning transformers and meticulous planning of contact points. The third aspect focuses on the sensing of contact through vision-based tactile sensors, as discussed in Chapter 8. This chapter presents a method for reconstructing contact profiles from image imprints captured by these sensors, providing a crucial feedback mechanism during manipulation tasks.Collectively, these contributions present a suite of novel grasping, manipulation, and assembly strategies. They are designed to reduce reliance on hand-engineering, thereby improving the efficiency and stability of robotic systems in diverse applications. This comprehensive body of work demonstrates the feasibility and benefits of integrating contact into robot learning. The effectiveness and practical applicability of these methods are rigorously validated through a series of simulations and real-world experiments involving various manipulators and hands.
■590 ▼aSchool code: 0028.
■650 4▼aRobotics
■650 4▼aComputer science
■650 4▼aMechanical engineering
■650 4▼aComputer engineering
■653 ▼aContact mechanics
■653 ▼aMachine learning
■653 ▼aRobotic grasping
■653 ▼aRobotic manipulation
■653 ▼aSensors
■690 ▼a0771
■690 ▼a0984
■690 ▼a0548
■690 ▼a0464
■690 ▼a0800
■71020▼aUniversity of California, Berkeley▼bMechanical Engineering.
■7730 ▼tDissertations Abstracts International▼g86-03B.
■790 ▼a0028
■791 ▼aPh.D.
■792 ▼a2024
■793 ▼aEnglish
■85640▼uhttp://www.riss.kr/pdu/ddodLink.do?id=T17160341▼nKERIS▼z이 자료의 원문은 한국교육학술정보원에서 제공합니다.
