서브메뉴
검색
Autonomous and Robust Monocular Simultaneous Localization and Mapping-Based Navigation for Robotic Operations in Space
Autonomous and Robust Monocular Simultaneous Localization and Mapping-Based Navigation for Robotic Operations in Space
상세정보
- 자료유형
- 학위논문 서양
- 최종처리일시
- 20260202105515
- ISBN
- 9798263342784
- DDC
- 531.14
- 저자명
- Dor, Mehregan.
- 서명/저자
- Autonomous and Robust Monocular Simultaneous Localization and Mapping-Based Navigation for Robotic Operations in Space
- 발행사항
- [Sl] : Georgia Institute of Technology, 2024
- 발행사항
- Ann Arbor : ProQuest Dissertations & Theses, 2024
- 형태사항
- 215 p
- 주기사항
- Source: Dissertations Abstracts International, Volume: 87-05, Section: B.
- 주기사항
- Advisor: Tsiotras, Panagiotis.
- 학위논문주기
- Thesis (Ph.D.)--Georgia Institute of Technology, 2024.
- 초록/해제
- 요약Non-cooperative unmapped space targets are typically poorly characterized natural or artificial orbiting bodies selected for inspection by means of an observer spacecraft circumnavigating it. We may include in space targets such things as resident space objects (RSO) and space debris, but equally we may designate unexplored small celestial bodies, such as comets and asteroids. We acknowledge the impact of autonomous navigation solutions for space applications in both rendezvous and proximity operations in Earth vicinity, as well as small body probing and surveying missions in deep space setting. The ever more capable on-board computation systems which allow high-rate, high data-flow applications to be run in real-time, provide the means for accurate navigation algorithms to also be run on-board the observer spacecraft, for the purpose of closed-loop control of the spacecraft's attitude and orbital motions. Yet, current navigation pipelines involve highly complex procedures, requiring ground-segment human intervention, especially since the target object's shape or dynamical properties are unknown before the encounter.Meanwhile, in ground robotics research, recent strides in autonomous and precise robotic localization and mapping, piggy-backed on the advances in computer vision-based procedures and the incorporation of machine learning techniques, have improved performance, accuracy, robustness and reliability by leaps and bounds, often leaving classical methods in the dust. We highlight Simultaenous Localization and Mapping (SLAM), which, with its visual (vSLAM) and visual-inertial (viSLAM) flavors, in both monocular and stereo vision settings, has produced impressive and efficient results for large scale mapping and localization tasks.Tomorrow's next generation of space-bound applications will certainly need to incorporate and adapt today's most recent solutions from the ground robotics community to further increase the breadth and scope of practicable space mission designs.In this dissertation, an attempt is made to bridge the gap between space and ground robotics when it comes to the state-of-the-art of precise navigation solutions. Specifically, the realm of real-time algorithms, capable of running on-the-fly and using a monocular vision-based measurement paradigm, are examined. By preferring monocular measurements, the work is contrasted to the most recent research efforts in spacecraft rendezvous and proximity operations, which typically favor stereo vision. SLAM is selected as a framework solution for model-agnostic non-cooperative navigation. A case study is conducted to determine the underlying real-world challenges of applying monocular SLAM to spacecraft relative navigation. The relevant constraints imposed by the relative orbital dynamics are leveraged to improve the accuracy of the navigation solution, instead of relying on uninformative inertial sensor accelerometer measurements. The proposed solution is developed within a state-of-the-art incremental smoothing-based estimation framework. Through validation and testing with simulation data, legacy mission data, and in-lab generated data, the accuracy and efficiency of the algorithms are demonstrated. An existing procedure for estimating the dynamical parameters of a spinning target spacecraft is extended to perform the estimation of spin state, center of mass and gravity parameter of a small body, on-the-fly. A modern technique, inspired by applications in mobile augmented reality, is exploited to tackle the challenges plaguing the estimation of a starting map and robot state for SLAM initialization, with assumption of weak-perspective projection and small camera baseline.
- 일반주제명
- Gravity
- 일반주제명
- Space telescopes
- 일반주제명
- Accelerometers
- 일반주제명
- Lie groups
- 일반주제명
- Parameter estimation
- 일반주제명
- Robotics
- 일반주제명
- Astronomy
- 일반주제명
- Mathematics
- 일반주제명
- Optics
- 기본자료저록
- Dissertations Abstracts International. 87-05B.
- 전자적 위치 및 접속
- 로그인 후 원문을 볼 수 있습니다.
MARC
008260126s2024 us c eng d■001000017360381
■00520260202105515
■006m o d
■007cr#unu||||||||
■020 ▼a9798263342784
■035 ▼a(MiAaPQ)AAI32309311
■035 ▼a(MiAaPQ)GeorgiaTech75319
■040 ▼aMiAaPQ▼cMiAaPQ
■0820 ▼a531.14
■1001 ▼aDor, Mehregan.
■24510▼aAutonomous and Robust Monocular Simultaneous Localization and Mapping-Based Navigation for Robotic Operations in Space
■260 ▼a[Sl]▼bGeorgia Institute of Technology▼c2024
■260 1▼aAnn Arbor▼bProQuest Dissertations & Theses▼c2024
■300 ▼a215 p
■500 ▼aSource: Dissertations Abstracts International, Volume: 87-05, Section: B.
■500 ▼aAdvisor: Tsiotras, Panagiotis.
■5021 ▼aThesis (Ph.D.)--Georgia Institute of Technology, 2024.
■520 ▼aNon-cooperative unmapped space targets are typically poorly characterized natural or artificial orbiting bodies selected for inspection by means of an observer spacecraft circumnavigating it. We may include in space targets such things as resident space objects (RSO) and space debris, but equally we may designate unexplored small celestial bodies, such as comets and asteroids. We acknowledge the impact of autonomous navigation solutions for space applications in both rendezvous and proximity operations in Earth vicinity, as well as small body probing and surveying missions in deep space setting. The ever more capable on-board computation systems which allow high-rate, high data-flow applications to be run in real-time, provide the means for accurate navigation algorithms to also be run on-board the observer spacecraft, for the purpose of closed-loop control of the spacecraft's attitude and orbital motions. Yet, current navigation pipelines involve highly complex procedures, requiring ground-segment human intervention, especially since the target object's shape or dynamical properties are unknown before the encounter.Meanwhile, in ground robotics research, recent strides in autonomous and precise robotic localization and mapping, piggy-backed on the advances in computer vision-based procedures and the incorporation of machine learning techniques, have improved performance, accuracy, robustness and reliability by leaps and bounds, often leaving classical methods in the dust. We highlight Simultaenous Localization and Mapping (SLAM), which, with its visual (vSLAM) and visual-inertial (viSLAM) flavors, in both monocular and stereo vision settings, has produced impressive and efficient results for large scale mapping and localization tasks.Tomorrow's next generation of space-bound applications will certainly need to incorporate and adapt today's most recent solutions from the ground robotics community to further increase the breadth and scope of practicable space mission designs.In this dissertation, an attempt is made to bridge the gap between space and ground robotics when it comes to the state-of-the-art of precise navigation solutions. Specifically, the realm of real-time algorithms, capable of running on-the-fly and using a monocular vision-based measurement paradigm, are examined. By preferring monocular measurements, the work is contrasted to the most recent research efforts in spacecraft rendezvous and proximity operations, which typically favor stereo vision. SLAM is selected as a framework solution for model-agnostic non-cooperative navigation. A case study is conducted to determine the underlying real-world challenges of applying monocular SLAM to spacecraft relative navigation. The relevant constraints imposed by the relative orbital dynamics are leveraged to improve the accuracy of the navigation solution, instead of relying on uninformative inertial sensor accelerometer measurements. The proposed solution is developed within a state-of-the-art incremental smoothing-based estimation framework. Through validation and testing with simulation data, legacy mission data, and in-lab generated data, the accuracy and efficiency of the algorithms are demonstrated. An existing procedure for estimating the dynamical parameters of a spinning target spacecraft is extended to perform the estimation of spin state, center of mass and gravity parameter of a small body, on-the-fly. A modern technique, inspired by applications in mobile augmented reality, is exploited to tackle the challenges plaguing the estimation of a starting map and robot state for SLAM initialization, with assumption of weak-perspective projection and small camera baseline.
■590 ▼aSchool code: 0078.
■650 4▼aGravity
■650 4▼aSpace telescopes
■650 4▼aAccelerometers
■650 4▼aLie groups
■650 4▼aParameter estimation
■650 4▼aRobotics
■650 4▼aAstronomy
■650 4▼aMathematics
■650 4▼aOptics
■690 ▼a0771
■690 ▼a0606
■690 ▼a0405
■690 ▼a0752
■71020▼aGeorgia Institute of Technology.
■7730 ▼tDissertations Abstracts International▼g87-05B.
■790 ▼a0078
■791 ▼aPh.D.
■792 ▼a2024
■793 ▼aEnglish
■85640▼uhttp://www.riss.kr/pdu/ddodLink.do?id=T17360381▼nKERIS▼z이 자료의 원문은 한국교육학술정보원에서 제공합니다.


