서브메뉴
검색
From Sensors to Intelligence: Adaptive Data Collection and Processing for Scalable Edge Systems
From Sensors to Intelligence: Adaptive Data Collection and Processing for Scalable Edge Systems
상세정보
- 자료유형
- 학위논문 서양
- 최종처리일시
- 20260202105202
- ISBN
- 9798273305366
- DDC
- 004
- 저자명
- Hu, Yi.
- 서명/저자
- From Sensors to Intelligence: Adaptive Data Collection and Processing for Scalable Edge Systems
- 발행사항
- [Sl] : Carnegie Mellon University, 2025
- 발행사항
- Ann Arbor : ProQuest Dissertations & Theses, 2025
- 형태사항
- 154 p
- 주기사항
- Source: Dissertations Abstracts International, Volume: 87-07, Section: B.
- 주기사항
- Advisor: Joe-Wong, Carlee;Iannucci, Bob.
- 학위논문주기
- Thesis (Ph.D.)--Carnegie Mellon University, 2025.
- 초록/해제
- 요약The proliferation of Internet of Things (IoT) devices and edge computing infrastructure transforms how we collect, process, and derive intelligence from sensor data. However, realizing this vision faces three fundamental bottlenecks: collecting data at scale under severe communication bandwidth and energy constraints, orchestrating computation across heterogeneous edge resources, and executing processing workloads on resource-constrained devices. This dissertation argues that overcoming these bottlenecks requires systematically exploiting structure---patterns and relationships that exist in sensor data, edge network resources, and processing workloads.We develop learning-based frameworks that automatically discover and leverage these structural properties across the data-to-intelligence pipeline. For scalable data collection, we introduce communication coordination systems that exploit temporal and spatial patterns in environmental sensor data. For intelligent resource orchestration, we present learning methods that utilize graph neural networks and Gaussian processes to capture contextual relationships between application workloads and edge resources. For efficient distributed execution, we develop a foundation model-powered correlated data analysis framework and a bit-width-adaptive quantized training method that tailors computation to device capabilities. Our frameworks enable collaborative data processing across heterogeneous edge devices.Through comprehensive evaluation on real-world datasets, wireless network simulators, and physical edge deployments, we demonstrate that structure-aware approaches enable scalable edge intelligence systems. Our contributions provide both practical solutions for deploying intelligent applications on resource-constrained edge infrastructure and principled frameworks for future research in edge computing, distributed learning, and networked systems.
- 일반주제명
- Computer science
- 일반주제명
- Systems science
- 일반주제명
- Computer engineering
- 키워드
- Edge computing
- 키워드
- Sensor data
- 키워드
- Wireless network
- 기타저자
- Carnegie Mellon University Electrical and Computer Engineering
- 기본자료저록
- Dissertations Abstracts International. 87-07B.
- 전자적 위치 및 접속
- 로그인 후 원문을 볼 수 있습니다.
MARC
008260126s2025 us c eng d■001000017359712
■00520260202105202
■006m o d
■007cr#unu||||||||
■020 ▼a9798273305366
■035 ▼a(MiAaPQ)AAI32245542
■040 ▼aMiAaPQ▼cMiAaPQ
■0820 ▼a004
■1001 ▼aHu, Yi.
■24510▼aFrom Sensors to Intelligence: Adaptive Data Collection and Processing for Scalable Edge Systems
■260 ▼a[Sl]▼bCarnegie Mellon University▼c2025
■260 1▼aAnn Arbor▼bProQuest Dissertations & Theses▼c2025
■300 ▼a154 p
■500 ▼aSource: Dissertations Abstracts International, Volume: 87-07, Section: B.
■500 ▼aAdvisor: Joe-Wong, Carlee;Iannucci, Bob.
■5021 ▼aThesis (Ph.D.)--Carnegie Mellon University, 2025.
■520 ▼aThe proliferation of Internet of Things (IoT) devices and edge computing infrastructure transforms how we collect, process, and derive intelligence from sensor data. However, realizing this vision faces three fundamental bottlenecks: collecting data at scale under severe communication bandwidth and energy constraints, orchestrating computation across heterogeneous edge resources, and executing processing workloads on resource-constrained devices. This dissertation argues that overcoming these bottlenecks requires systematically exploiting structure---patterns and relationships that exist in sensor data, edge network resources, and processing workloads.We develop learning-based frameworks that automatically discover and leverage these structural properties across the data-to-intelligence pipeline. For scalable data collection, we introduce communication coordination systems that exploit temporal and spatial patterns in environmental sensor data. For intelligent resource orchestration, we present learning methods that utilize graph neural networks and Gaussian processes to capture contextual relationships between application workloads and edge resources. For efficient distributed execution, we develop a foundation model-powered correlated data analysis framework and a bit-width-adaptive quantized training method that tailors computation to device capabilities. Our frameworks enable collaborative data processing across heterogeneous edge devices.Through comprehensive evaluation on real-world datasets, wireless network simulators, and physical edge deployments, we demonstrate that structure-aware approaches enable scalable edge intelligence systems. Our contributions provide both practical solutions for deploying intelligent applications on resource-constrained edge infrastructure and principled frameworks for future research in edge computing, distributed learning, and networked systems.
■590 ▼aSchool code: 0041.
■650 4▼aComputer science
■650 4▼aSystems science
■650 4▼aComputer engineering
■653 ▼aEdge computing
■653 ▼aSensor data
■653 ▼aWireless network
■653 ▼aGaussian processes
■653 ▼aGraph neural networks
■690 ▼a0984
■690 ▼a0790
■690 ▼a0464
■71020▼aCarnegie Mellon University▼bElectrical and Computer Engineering.
■7730 ▼tDissertations Abstracts International▼g87-07B.
■790 ▼a0041
■791 ▼aPh.D.
■792 ▼a2025
■793 ▼aEnglish
■85640▼uhttp://www.riss.kr/pdu/ddodLink.do?id=T17359712▼nKERIS▼z이 자료의 원문은 한국교육학술정보원에서 제공합니다.
