본문

서브메뉴

Methods for Contact-Rich Robot Manipulation
Methods for Contact-Rich Robot Manipulation
Methods for Contact-Rich Robot Manipulation

Detailed Information

자료유형  
 학위논문 서양
최종처리일시  
20260202104854
ISBN  
9798288815379
DDC  
620
저자명  
Chen, Claire.
서명/저자  
Methods for Contact-Rich Robot Manipulation
발행사항  
[Sl] : Stanford University, 2025
발행사항  
Ann Arbor : ProQuest Dissertations & Theses, 2025
형태사항  
77 p
주기사항  
Source: Dissertations Abstracts International, Volume: 87-02, Section: B.
주기사항  
Advisor: Bohg, Jeannette.
학위논문주기  
Thesis (Ph.D.)--Stanford University, 2025.
초록/해제  
요약To achieve human-level dexterity, robots must master contact-rich interactions with the physical world. The crux of contact-rich manipulation is making contact in the right places, at the right times, and with the right forces. Advancing contact-rich robot manipulation requires designing methods that focus on these most fundamental elements of contact. Towards this claim, this thesis presents novel grasp prediction, trajectory optimization, and imitation learning methods designed specifically for contact-rich interactions. We show that, compared to baselines, our methods improve the contact-rich manipulation capabilities of robots because they explicitly consider how to make contact in the right places, at the right times, and with the right forces. First, we consider the contact-rich task of interacting with articulated objects, where robot motion is constrained by the moving parts of objects. We design a grasp prediction model that centers on making contact with articulated objects in the right places. We show how combining our grasp prediction model with compliant control enables robots to robustly interact with articulated objects that have varied local geometries, articulation axes, and joint states. Next, we consider dexterous manipulation, where robots must apply precise and coordinated forces across multiple contact points. Within this setting, we explore both analytical and data- driven methods. On the analytical side, we show how discrete contact planning can be used to enforce making contact at the right places and times within contact-implicit trajectory optimization (CITO). Our method plans more robust dexterous manipulation trajectories than a general CITO baseline. On the data-driven side, we present a method that leverages force sensing to compute robot actions for policy learning that apply the right forces. These force-informed actions improve policy learning for precise and coordinated dexterous manipulation tasks like opening an AirPods case and unscrewing a nut. Together, these contributions show how tailoring grasp prediction, trajectory optimization, and imitation learning methods to focus on precise contact brings robots closer to human-level dexterity. Still, many challenges remain for robots to achieve the same levels of generalization and robustness as people, particularly in contact-rich settings. To conclude, we share some future directions for continuing to improve contact-rich robot manipulation.
일반주제명  
Robots
일반주제명  
Teaching methods
일반주제명  
Success
일반주제명  
Planning
일반주제명  
Computer science
일반주제명  
Computer engineering
일반주제명  
Robotics
키워드  
Contact-implicit trajectory optimization
키워드  
Contact-rich interactions
키워드  
Dexterous manipulation tasks
기타저자  
Stanford University.
기본자료저록  
Dissertations Abstracts International. 87-02B.
전자적 위치 및 접속  
로그인 후 원문을 볼 수 있습니다.

MARC

 008260126s2025        us                              c    eng  d
■001000017359240
■00520260202104854
■006m          o    d                
■007cr#unu||||||||
■020    ▼a9798288815379
■035    ▼a(MiAaPQ)AAI32200992
■035    ▼a(MiAaPQ)Stanfordsh608vx1165
■040    ▼aMiAaPQ▼cMiAaPQ
■0820  ▼a620
■1001  ▼aChen,  Claire.
■24510▼aMethods  for  Contact-Rich  Robot  Manipulation
■260    ▼a[Sl]▼bStanford  University▼c2025
■260  1▼aAnn  Arbor▼bProQuest  Dissertations  &  Theses▼c2025
■300    ▼a77  p
■500    ▼aSource:  Dissertations  Abstracts  International,  Volume:  87-02,  Section:  B.
■500    ▼aAdvisor:  Bohg,  Jeannette.
■5021  ▼aThesis  (Ph.D.)--Stanford  University,  2025.
■520    ▼aTo  achieve  human-level  dexterity,  robots  must  master  contact-rich  interactions  with  the  physical  world.  The  crux  of  contact-rich  manipulation  is  making  contact  in  the  right  places,  at  the  right  times,  and  with  the  right  forces.  Advancing  contact-rich  robot  manipulation  requires  designing  methods  that  focus  on  these  most  fundamental  elements  of  contact.  Towards  this  claim,  this  thesis  presents  novel  grasp  prediction,  trajectory  optimization,  and  imitation  learning  methods  designed  specifically  for  contact-rich  interactions.  We  show  that,  compared  to  baselines,  our  methods  improve  the  contact-rich  manipulation  capabilities  of  robots  because  they  explicitly  consider  how  to  make  contact  in  the  right  places,  at  the  right  times,  and  with  the  right  forces.  First,  we  consider  the  contact-rich  task  of  interacting  with  articulated  objects,  where  robot  motion  is  constrained  by  the  moving  parts  of  objects.  We  design  a  grasp  prediction  model  that  centers  on  making  contact  with  articulated  objects  in  the  right  places.  We  show  how  combining  our  grasp  prediction  model  with  compliant  control  enables  robots  to  robustly  interact  with  articulated  objects  that  have  varied  local  geometries,  articulation  axes,  and  joint  states.  Next,  we  consider  dexterous  manipulation,  where  robots  must  apply  precise  and  coordinated  forces  across  multiple  contact  points.  Within  this  setting,  we  explore  both  analytical  and  data-  driven  methods.  On  the  analytical  side,  we  show  how  discrete  contact  planning  can  be  used  to  enforce  making  contact  at  the  right  places  and  times  within  contact-implicit  trajectory  optimization  (CITO).  Our  method  plans  more  robust  dexterous  manipulation  trajectories  than  a  general  CITO  baseline.  On  the  data-driven  side,  we  present  a  method  that  leverages  force  sensing  to  compute  robot  actions  for  policy  learning  that  apply  the  right  forces.  These  force-informed  actions  improve  policy  learning  for  precise  and  coordinated  dexterous  manipulation  tasks  like  opening  an  AirPods  case  and  unscrewing  a  nut.  Together,  these  contributions  show  how  tailoring  grasp  prediction,  trajectory  optimization,  and  imitation  learning  methods  to  focus  on  precise  contact  brings  robots  closer  to  human-level  dexterity.  Still,  many  challenges  remain  for  robots  to  achieve  the  same  levels  of  generalization  and  robustness  as  people,  particularly  in  contact-rich  settings.  To  conclude,  we  share  some  future  directions  for  continuing  to  improve  contact-rich  robot  manipulation.
■590    ▼aSchool  code:  0212.
■650  4▼aRobots
■650  4▼aTeaching  methods
■650  4▼aSuccess
■650  4▼aPlanning
■650  4▼aComputer  science
■650  4▼aComputer  engineering
■650  4▼aRobotics
■653    ▼aContact-implicit  trajectory  optimization
■653    ▼aContact-rich  interactions
■653    ▼aDexterous  manipulation  tasks
■690    ▼a0771
■690    ▼a0984
■690    ▼a0464
■71020▼aStanford  University.
■7730  ▼tDissertations  Abstracts  International▼g87-02B.
■790    ▼a0212
■791    ▼aPh.D.
■792    ▼a2025
■793    ▼aEnglish
■85640▼uhttp://www.riss.kr/pdu/ddodLink.do?id=T17359240▼nKERIS▼z이  자료의  원문은  한국교육학술정보원에서  제공합니다.

Preview

Export

ChatGPT Discussion

AI Recommended Related Books


    New Books MORE
    Statistics for the past 3 years. Go to brief

    Подробнее информация.

    • Бронирование
    • не существует
    • моя папка
    • Первый запрос зрения
    • Non-Book Loan Application
    • Nighttime Book Loan Application
    материал
    Reg No. Количество платежных Местоположение статус Ленд информации
    TF14911 전자도서 대출가능 My Folder 부재도서신고 비도서대출신청 야간 도서대출신청

    * Бронирование доступны в заимствований книги. Чтобы сделать предварительный заказ, пожалуйста, нажмите кнопку бронирование

    Books borrowed together with this book

    Related Popular Books

    Available after logging in.