서브메뉴
검색
Safe Planning Under Uncertainty Using Surrogate Models
Safe Planning Under Uncertainty Using Surrogate Models
상세정보
- 자료유형
- 학위논문 서양
- 최종처리일시
- 20260202104747
- ISBN
- 9798290652283
- DDC
- 150
- 서명/저자
- Safe Planning Under Uncertainty Using Surrogate Models
- 발행사항
- [Sl] : Stanford University, 2025
- 발행사항
- Ann Arbor : ProQuest Dissertations & Theses, 2025
- 형태사항
- 201 p
- 주기사항
- Source: Dissertations Abstracts International, Volume: 87-01, Section: B.
- 주기사항
- Advisor: Barrett, Clark;Kochenderfer, Mykel.
- 학위논문주기
- Thesis (Ph.D.)--Stanford University, 2025.
- 초록/해제
- 요약To make safe decisions in real-world environments, algorithms must account for the inherent uncertainty in perception systems and agent dynamics, resulting in high-dimensional problems. The use of surrogate models to replace hand-crafted planning heuristics and avoid running the computationally expensive true system has shown promise in enabling large-scale, safe planning. This thesis introduces five main contributions to address the challenges of safe planning under uncertainty and validation. To improve planning efficiency over beliefs in partially observable Markov decision processes (POMDPs), we introduce batched belief-stateMDPs, which abstract belief-state planning using parallelizable batches of the underlying models. This abstraction requires models that can be easily parallelized; therefore, we can learn surrogate transition and observation models from data and propose the inversion variational autoencoder(I-VAE) to sample from the posterior belief given partial observations. To replace planning heuristics and enable long-horizon planning, we introduce BetaZero, a policy iteration algorithm that combines offline learning with online belief-state planning. Extending BetaZero to safety-critical problems, we propose ConstrainedZero, which solves chance-constrained POMDPs by optimizing the balance between utility and a target level of safety. Finally, given a learned safe policy, we develop a Bayesian safety validationmethod to estimate the failure probability of a black-box system using probabilistic surrogate models. We apply these algorithms to real-world problems, including aircraft collision avoidance, autonomous aircraft runway detection, safe carbon capture and storage, critical mineral exploration, robot navigation, and aerial wildfire suppression.
- 일반주제명
- Success
- 일반주제명
- Robots
- 일반주제명
- Adaptation
- 일반주제명
- Aviation
- 일반주제명
- Collisions
- 일반주제명
- Climate change
- 일반주제명
- Pilots
- 일반주제명
- Robotics
- 일반주제명
- Geology
- 일반주제명
- Aircraft
- 일반주제명
- Carbon sequestration
- 일반주제명
- Failure analysis
- 일반주제명
- Planning
- 일반주제명
- Decision making
- 일반주제명
- Neural networks
- 일반주제명
- Design
- 일반주제명
- Probability
- 기타저자
- Stanford University.
- 기본자료저록
- Dissertations Abstracts International. 87-01B.
- 전자적 위치 및 접속
- 로그인 후 원문을 볼 수 있습니다.
MARC
008260126s2025 us c eng d■001000017358753
■00520260202104747
■006m o d
■007cr#unu||||||||
■020 ▼a9798290652283
■035 ▼a(MiAaPQ)AAI32149762
■035 ▼a(MiAaPQ)Stanfordzg643rb0595
■040 ▼aMiAaPQ▼cMiAaPQ
■0820 ▼a150
■1001 ▼aMoss, Robert John.
■24510▼aSafe Planning Under Uncertainty Using Surrogate Models
■260 ▼a[Sl]▼bStanford University▼c2025
■260 1▼aAnn Arbor▼bProQuest Dissertations & Theses▼c2025
■300 ▼a201 p
■500 ▼aSource: Dissertations Abstracts International, Volume: 87-01, Section: B.
■500 ▼aAdvisor: Barrett, Clark;Kochenderfer, Mykel.
■5021 ▼aThesis (Ph.D.)--Stanford University, 2025.
■520 ▼aTo make safe decisions in real-world environments, algorithms must account for the inherent uncertainty in perception systems and agent dynamics, resulting in high-dimensional problems. The use of surrogate models to replace hand-crafted planning heuristics and avoid running the computationally expensive true system has shown promise in enabling large-scale, safe planning. This thesis introduces five main contributions to address the challenges of safe planning under uncertainty and validation. To improve planning efficiency over beliefs in partially observable Markov decision processes (POMDPs), we introduce batched belief-stateMDPs, which abstract belief-state planning using parallelizable batches of the underlying models. This abstraction requires models that can be easily parallelized; therefore, we can learn surrogate transition and observation models from data and propose the inversion variational autoencoder(I-VAE) to sample from the posterior belief given partial observations. To replace planning heuristics and enable long-horizon planning, we introduce BetaZero, a policy iteration algorithm that combines offline learning with online belief-state planning. Extending BetaZero to safety-critical problems, we propose ConstrainedZero, which solves chance-constrained POMDPs by optimizing the balance between utility and a target level of safety. Finally, given a learned safe policy, we develop a Bayesian safety validationmethod to estimate the failure probability of a black-box system using probabilistic surrogate models. We apply these algorithms to real-world problems, including aircraft collision avoidance, autonomous aircraft runway detection, safe carbon capture and storage, critical mineral exploration, robot navigation, and aerial wildfire suppression.
■590 ▼aSchool code: 0212.
■650 4▼aSuccess
■650 4▼aRobots
■650 4▼aAdaptation
■650 4▼aAviation
■650 4▼aCollisions
■650 4▼aClimate change
■650 4▼aPilots
■650 4▼aRobotics
■650 4▼aGeology
■650 4▼aAircraft
■650 4▼aCarbon sequestration
■650 4▼aFailure analysis
■650 4▼aPlanning
■650 4▼aDecision making
■650 4▼aNeural networks
■650 4▼aDesign
■650 4▼aProbability
■690 ▼a0771
■690 ▼a0372
■690 ▼a0389
■690 ▼a0404
■71020▼aStanford University.
■7730 ▼tDissertations Abstracts International▼g87-01B.
■790 ▼a0212
■791 ▼aPh.D.
■792 ▼a2025
■793 ▼aEnglish
■85640▼uhttp://www.riss.kr/pdu/ddodLink.do?id=T17358753▼nKERIS▼z이 자료의 원문은 한국교육학술정보원에서 제공합니다.


