서브메뉴
검색
Multi-Future Environment Forecasting and Occlusion Inference for Autonomous Driving
Multi-Future Environment Forecasting and Occlusion Inference for Autonomous Driving
상세정보
- 자료유형
- 학위논문 서양
- 최종처리일시
- 20260202105626
- ISBN
- 9798265429681
- DDC
- 001
- 저자명
- Lange, Bernard.
- 서명/저자
- Multi-Future Environment Forecasting and Occlusion Inference for Autonomous Driving
- 발행사항
- [Sl] : Stanford University, 2025
- 발행사항
- Ann Arbor : ProQuest Dissertations & Theses, 2025
- 형태사항
- 109 p
- 주기사항
- Source: Dissertations Abstracts International, Volume: 87-05, Section: A.
- 주기사항
- Advisor: Kochenderfer, Mykel.
- 학위논문주기
- Thesis (Ph.D.)--Stanford University, 2025.
- 초록/해제
- 요약Human-operated vehicles play a critical role in modern society by enabling the efficient movement of people and goods, thereby saving time and fostering economic growth. Despite numerous technical improvements in vehicle and road safety, they reveal human limitations in operating these vehicles. These challenges have motivated research into self-driving cars. Although today's autonomous vehicles operate reliably in well-mapped, geofenced areas, they struggle to generalize to complex situations that demand deep scene understanding and reasoning about many interacting and often occluded agents. Such failures highlight the need for multi-future environment prediction methods, which are the focus of this thesis.This dissertation presents learning-based techniques for predicting the environment surrounding an autonomous vehicle, using information ranging from raw sensor measurements to vectorized representations provided by the perception module. We begin by addressing the full-observability assumption common in trajectory prediction models. In the first half of this thesis, we introduce a framework capable of jointly performing trajectory prediction and occlusion inference under partial observability, yielding improved robustness and state-of-the-art occlusion inference.A key limitation of vectorized methods is their reliance on manually labeled data and the marginalized prediction setting. To overcome these limitations, we develop a sensor-space approach that predicts LiDAR-based occupancy grid maps while conditioning on complementary modalities such as RGB cameras, planned trajectories, and maps. Our approach generates multiple hypotheses of future occupancy sequences in real time and achieves state-of-the-art performance, with clear gains from multimodal conditioning.
- 일반주제명
- Visualization
- 일반주제명
- Autonomous vehicles
- 일반주제명
- Robotics
- 일반주제명
- Transportation
- 기타저자
- Stanford University.
- 기본자료저록
- Dissertations Abstracts International. 87-05A.
- 전자적 위치 및 접속
- 로그인 후 원문을 볼 수 있습니다.
MARC
008260126s2025 us c eng d■001000017360835
■00520260202105626
■006m o d
■007cr#unu||||||||
■020 ▼a9798265429681
■035 ▼a(MiAaPQ)AAI32316554
■035 ▼a(MiAaPQ)Stanfordnq917qm4849
■040 ▼aMiAaPQ▼cMiAaPQ
■0820 ▼a001
■1001 ▼aLange, Bernard.
■24510▼aMulti-Future Environment Forecasting and Occlusion Inference for Autonomous Driving
■260 ▼a[Sl]▼bStanford University▼c2025
■260 1▼aAnn Arbor▼bProQuest Dissertations & Theses▼c2025
■300 ▼a109 p
■500 ▼aSource: Dissertations Abstracts International, Volume: 87-05, Section: A.
■500 ▼aAdvisor: Kochenderfer, Mykel.
■5021 ▼aThesis (Ph.D.)--Stanford University, 2025.
■520 ▼aHuman-operated vehicles play a critical role in modern society by enabling the efficient movement of people and goods, thereby saving time and fostering economic growth. Despite numerous technical improvements in vehicle and road safety, they reveal human limitations in operating these vehicles. These challenges have motivated research into self-driving cars. Although today's autonomous vehicles operate reliably in well-mapped, geofenced areas, they struggle to generalize to complex situations that demand deep scene understanding and reasoning about many interacting and often occluded agents. Such failures highlight the need for multi-future environment prediction methods, which are the focus of this thesis.This dissertation presents learning-based techniques for predicting the environment surrounding an autonomous vehicle, using information ranging from raw sensor measurements to vectorized representations provided by the perception module. We begin by addressing the full-observability assumption common in trajectory prediction models. In the first half of this thesis, we introduce a framework capable of jointly performing trajectory prediction and occlusion inference under partial observability, yielding improved robustness and state-of-the-art occlusion inference.A key limitation of vectorized methods is their reliance on manually labeled data and the marginalized prediction setting. To overcome these limitations, we develop a sensor-space approach that predicts LiDAR-based occupancy grid maps while conditioning on complementary modalities such as RGB cameras, planned trajectories, and maps. Our approach generates multiple hypotheses of future occupancy sequences in real time and achieves state-of-the-art performance, with clear gains from multimodal conditioning.
■590 ▼aSchool code: 0212.
■650 4▼aVisualization
■650 4▼aAutonomous vehicles
■650 4▼aRobotics
■650 4▼aTransportation
■690 ▼a0771
■690 ▼a0709
■71020▼aStanford University.
■7730 ▼tDissertations Abstracts International▼g87-05A.
■790 ▼a0212
■791 ▼aPh.D.
■792 ▼a2025
■793 ▼aEnglish
■85640▼uhttp://www.riss.kr/pdu/ddodLink.do?id=T17360835▼nKERIS▼z이 자료의 원문은 한국교육학술정보원에서 제공합니다.
