본문

서브메뉴

Safe Decision-Making for Responsible Robotics
Safe Decision-Making for Responsible Robotics
Safe Decision-Making for Responsible Robotics

Detailed Information

자료유형  
 학위논문 서양
최종처리일시  
20260202104722
ISBN  
9798290938622
DDC  
629.8
저자명  
Lyu, Yiwei.
서명/저자  
Safe Decision-Making for Responsible Robotics
발행사항  
[Sl] : Carnegie Mellon University, 2025
발행사항  
Ann Arbor : ProQuest Dissertations & Theses, 2025
형태사항  
98 p
주기사항  
Source: Dissertations Abstracts International, Volume: 87-02, Section: B.
주기사항  
Advisor: Dolan, John.
학위논문주기  
Thesis (Ph.D.)--Carnegie Mellon University, 2025.
초록/해제  
요약As robots become integrated into modern society and play key roles in industries like transportation, healthcare, and manufacturing, these systems must interact closely with humans and make real-time decisions that prioritize safety. This growing reliance highlights the need for responsible robotics-systems that not only perform tasks efficiently but also make reliable, safe, and ethically sound decisions in complex environments. The most critical aspect of responsible robotics is ensuring physical safety, preventing harm during operation. However, achieving this is not as simple as telling robots to "avoid collisions", as robots may face sensing failures, environmental disturbances, and unpredictable behaviors from others. Drawing inspiration from control theory, machine learning, and social science, this thesis aims to develop theories and algorithms that enable robots to proactively identify potential risks, mitigate them effectively, and verify their decisions in real time to ensure safe operation throughout the execution of their tasks in complex environments while balancing safety and efficiency.This thesis focuses on developing robot decision-making methodologies with formal physical safety guarantees. My first line of work addresses the functional safety of individual robots' actions, ensuring they can operate robustly against motion uncertainties. A probabilistic adaptive safety verification filter ensures immediate safety and a risk evaluation framework biases robot decisions to account for long-term safety. This allows robots to adapt to varying levels of uncertainty, displaying a range of safety-critical behaviors from conservative to efficient. The concept of functional safety is then extended to address collaborative safety in multi-robot systems, where teams of robots must interact safely and efficiently. The thesis formalizes the first responsibility-reasoning framework, allowing individual robots to contribute to a collective goal by reasoning over their roles in a decentralized manner. This framework demonstrates group intelligence, allowing robots with diverse capabilities and intentions to collaborate effectively with aligned heterogeneity while maintaining both safety and efficiency. Lastly, we address the challenge of representational safety, which emerges when safety specifications either fail to align with social expectations or are not explicitly defined in advance. A risk-informed specification formalization contextualizes high-level moral principles, such as those inspired by Asimov's Laws, into physically meaningful motion constraints, enabling robots to exhibit socially compliant behavior. For under-defined safety specifications, a causality-embedded, model-based learning approach infers the unknown parameters of implicit safety notions from interaction data and behavioral observations. Together, these methods enable robots to uphold safety concepts grounded in human values, while learning ambiguous or evolving safety specifications without requiring explicit labels.Collectively, these efforts support the broader goal of responsible robotics by enabling robots to reason about and respond to external environments at multiple levels, from individual decision-making to team coordination and specification learning. By advancing safety in function, collaboration, and representation, this thesis lays the groundwork for autonomous systems that are not only physically safe, but also capable of behaving responsibly in uncertain and dynamic environments.
일반주제명  
Robotics
일반주제명  
Computer engineering
일반주제명  
Electrical engineering
키워드  
Behavior planning
키워드  
Multi-robot systems
키워드  
Safe control and verification
기타저자  
Carnegie Mellon University Electrical and Computer Engineering
기본자료저록  
Dissertations Abstracts International. 87-02B.
전자적 위치 및 접속  
로그인 후 원문을 볼 수 있습니다.

MARC

 008260126s2025        us                              c    eng  d
■001000017358579
■00520260202104722
■006m          o    d                
■007cr#unu||||||||
■020    ▼a9798290938622
■035    ▼a(MiAaPQ)AAI32121574
■040    ▼aMiAaPQ▼cMiAaPQ
■0820  ▼a629.8
■1001  ▼aLyu,  Yiwei.▼0(orcid)0009-0007-0136-5369
■24510▼aSafe  Decision-Making  for  Responsible  Robotics
■260    ▼a[Sl]▼bCarnegie  Mellon  University▼c2025
■260  1▼aAnn  Arbor▼bProQuest  Dissertations  &  Theses▼c2025
■300    ▼a98  p
■500    ▼aSource:  Dissertations  Abstracts  International,  Volume:  87-02,  Section:  B.
■500    ▼aAdvisor:  Dolan,  John.
■5021  ▼aThesis  (Ph.D.)--Carnegie  Mellon  University,  2025.
■520    ▼aAs  robots  become  integrated  into  modern  society  and  play  key  roles  in  industries  like  transportation,  healthcare,  and  manufacturing,  these  systems  must  interact  closely  with  humans  and  make  real-time  decisions  that  prioritize  safety.  This  growing  reliance  highlights  the  need  for  responsible  robotics-systems  that  not  only  perform  tasks  efficiently  but  also  make  reliable,  safe,  and  ethically  sound  decisions  in  complex  environments.  The  most  critical  aspect  of  responsible  robotics  is  ensuring  physical  safety,  preventing  harm  during  operation.  However,  achieving  this  is  not  as  simple  as  telling  robots  to  "avoid  collisions",  as  robots  may  face  sensing  failures,  environmental  disturbances,  and  unpredictable  behaviors  from  others.  Drawing  inspiration  from  control  theory,  machine  learning,  and  social  science,  this  thesis  aims  to  develop  theories  and  algorithms  that  enable  robots  to  proactively  identify  potential  risks,  mitigate  them  effectively,  and  verify  their  decisions  in  real  time  to  ensure  safe  operation  throughout  the  execution  of  their  tasks  in  complex  environments  while  balancing  safety  and  efficiency.This  thesis  focuses  on  developing  robot  decision-making  methodologies  with  formal  physical  safety  guarantees.  My  first  line  of  work  addresses  the  functional  safety  of  individual  robots'  actions,  ensuring  they  can  operate  robustly  against  motion  uncertainties.  A  probabilistic  adaptive  safety  verification  filter  ensures  immediate  safety  and  a  risk  evaluation  framework  biases  robot  decisions  to  account  for  long-term  safety.  This  allows  robots  to  adapt  to  varying  levels  of  uncertainty,  displaying  a  range  of  safety-critical  behaviors  from  conservative  to  efficient.  The  concept  of  functional  safety  is  then  extended  to  address  collaborative  safety  in  multi-robot  systems,  where  teams  of  robots  must  interact  safely  and  efficiently.  The  thesis  formalizes  the  first  responsibility-reasoning  framework,  allowing  individual  robots  to  contribute  to  a  collective  goal  by  reasoning  over  their  roles  in  a  decentralized  manner.  This  framework  demonstrates  group  intelligence,  allowing  robots  with  diverse  capabilities  and  intentions  to  collaborate  effectively  with  aligned  heterogeneity  while  maintaining  both  safety  and  efficiency.  Lastly,  we  address  the  challenge  of  representational  safety,  which  emerges  when  safety  specifications  either  fail  to  align  with  social  expectations  or  are  not  explicitly  defined  in  advance.  A  risk-informed  specification  formalization  contextualizes  high-level  moral  principles,  such  as  those  inspired  by  Asimov's  Laws,  into  physically  meaningful  motion  constraints,  enabling  robots  to  exhibit  socially  compliant  behavior.  For  under-defined  safety  specifications,  a  causality-embedded,  model-based  learning  approach  infers  the  unknown  parameters  of  implicit  safety  notions  from  interaction  data  and  behavioral  observations.  Together,  these  methods  enable  robots  to  uphold  safety  concepts  grounded  in  human  values,  while  learning  ambiguous  or  evolving  safety  specifications  without  requiring  explicit  labels.Collectively,  these  efforts  support  the  broader  goal  of  responsible  robotics  by  enabling  robots  to  reason  about  and  respond  to  external  environments  at  multiple  levels,  from  individual  decision-making  to  team  coordination  and  specification  learning.  By  advancing  safety  in  function,  collaboration,  and  representation,  this  thesis  lays  the  groundwork  for  autonomous  systems  that  are  not  only  physically  safe,  but  also  capable  of  behaving  responsibly  in  uncertain  and  dynamic  environments.
■590    ▼aSchool  code:  0041.
■650  4▼aRobotics
■650  4▼aComputer  engineering
■650  4▼aElectrical  engineering
■653    ▼aBehavior  planning
■653    ▼aMulti-robot  systems
■653    ▼aSafe  control  and  verification
■690    ▼a0771
■690    ▼a0544
■690    ▼a0464
■71020▼aCarnegie  Mellon  University▼bElectrical  and  Computer  Engineering.
■7730  ▼tDissertations  Abstracts  International▼g87-02B.
■790    ▼a0041
■791    ▼aPh.D.
■792    ▼a2025
■793    ▼aEnglish
■85640▼uhttp://www.riss.kr/pdu/ddodLink.do?id=T17358579▼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
    TF18337 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.