서브메뉴
검색
Forecasting and Optimization of Urban Traffic for Improved Adaptation to Climate Change
Forecasting and Optimization of Urban Traffic for Improved Adaptation to Climate Change
상세정보
- 자료유형
- 학위논문 서양
- 최종처리일시
- 20260202105628
- ISBN
- 9798265429964
- DDC
- 388.4
- 저자명
- Chen, Yirong.
- 서명/저자
- Forecasting and Optimization of Urban Traffic for Improved Adaptation to Climate Change
- 발행사항
- [Sl] : Stanford University, 2025
- 발행사항
- Ann Arbor : ProQuest Dissertations & Theses, 2025
- 형태사항
- 165 p
- 주기사항
- Source: Dissertations Abstracts International, Volume: 87-05, Section: A.
- 주기사항
- Advisor: Lepech, Michael.
- 학위논문주기
- Thesis (Ph.D.)--Stanford University, 2025.
- 초록/해제
- 요약The rapid urbanization and exponential growth in vehicle ownership have exacerbated traffic congestion, a challenge further exacerbated by climate change-induced extreme weather events. These events disrupt traffic flow and amplify greenhouse gas emissions, posing significant barriers to efficient urban mobility. Climate change projections indicate increasing frequency and severity of extreme weather events, including heatwaves, heavy rainfall, and storms, all of which directly impair transportation by reducing road capacity, elevating accident risks, and altering travel behaviors. Conventional traffic management strategies, however, often fail to account for these climate-related disruptions systematically. At the same time, existing optimization frameworks, particularly reinforcement learning (RL)-based traffic signal control systems, assume instantaneous access to real-time data and rely on high-frequency interventions (e.g., every 5 seconds). This idealized setup conflicts with real-world constraints such as communication delays, computational resource limitations, and safety regulations that necessitate longer control intervals. As a result, current approaches struggle to maintain efficiency under dynamic weather conditions or operational constraints, highlighting a critical gap between research and practical implementation.To address these challenges, this dissertation introduces a holistic framework that integrates climate impact quantification, traffic forecasting, and adaptive optimization. First, we develop a comprehensive model to assess the environmental, economic, and social consequences of climate-induced traffic disruptions under three top-priority climate change scenarios. Our analysis reveals significant welfare losses due to prolonged congestion and increased emissions under moderate to high climate change scenarios, underscoring the urgency of proactive adaptation. Building on this foundation, we present three key contributions. The first is AHSTN (Adaptive Hierarchical Spatio-Temporal Network), an end-to-end graph convolutional network that dynamically captures multiscale spatiotemporal dependencies in traffic data. By leveraging hierarchical learning mechanisms, AHSTN achieves superior prediction accuracy while reducing computational complexity, enabling real-time forecasting of weather-sensitive traffic patterns. The second contribution is prediction-driven signal control, a novel framework that integrates traffic flow forecasts with real-time traffic data to optimize signal timings. This method dynamically adjusts phase durations by incorporating both current and predicted vehicle queues, ensuring robustness under communication delays and uncertain weather conditions. The third contribution is CCDA (Centralized Critic and Decentralized Actors), a scalable RL architecture designed for low-frequency signal control. CCDA introduces the adjust all phases action design, allowing simultaneous optimization of signal durations across all phases in a cycle, thereby maximizing the impact of each intervention. Through decentralized actors and a centralized critic, CCDA balances localized decision-making with global coordination, ensuring adaptability to varying control frequencies while maintaining computational efficiency. Collectively, these advancements provide a scalable and adaptive approach to urban traffic management, bridging the gap between theoretical models and real-world operational requirements while enhancing resilience to climate change.
- 일반주제명
- Traffic flow
- 일반주제명
- Commuting
- 일반주제명
- Climate change
- 일반주제명
- Rain
- 일반주제명
- Traffic control
- 일반주제명
- Meteorology
- 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■001000017360850
■00520260202105628
■006m o d
■007cr#unu||||||||
■020 ▼a9798265429964
■035 ▼a(MiAaPQ)AAI32316575
■035 ▼a(MiAaPQ)Stanfordqp135gh6053
■040 ▼aMiAaPQ▼cMiAaPQ
■0820 ▼a388.4
■1001 ▼aChen, Yirong.
■24510▼aForecasting and Optimization of Urban Traffic for Improved Adaptation to Climate Change
■260 ▼a[Sl]▼bStanford University▼c2025
■260 1▼aAnn Arbor▼bProQuest Dissertations & Theses▼c2025
■300 ▼a165 p
■500 ▼aSource: Dissertations Abstracts International, Volume: 87-05, Section: A.
■500 ▼aAdvisor: Lepech, Michael.
■5021 ▼aThesis (Ph.D.)--Stanford University, 2025.
■520 ▼aThe rapid urbanization and exponential growth in vehicle ownership have exacerbated traffic congestion, a challenge further exacerbated by climate change-induced extreme weather events. These events disrupt traffic flow and amplify greenhouse gas emissions, posing significant barriers to efficient urban mobility. Climate change projections indicate increasing frequency and severity of extreme weather events, including heatwaves, heavy rainfall, and storms, all of which directly impair transportation by reducing road capacity, elevating accident risks, and altering travel behaviors. Conventional traffic management strategies, however, often fail to account for these climate-related disruptions systematically. At the same time, existing optimization frameworks, particularly reinforcement learning (RL)-based traffic signal control systems, assume instantaneous access to real-time data and rely on high-frequency interventions (e.g., every 5 seconds). This idealized setup conflicts with real-world constraints such as communication delays, computational resource limitations, and safety regulations that necessitate longer control intervals. As a result, current approaches struggle to maintain efficiency under dynamic weather conditions or operational constraints, highlighting a critical gap between research and practical implementation.To address these challenges, this dissertation introduces a holistic framework that integrates climate impact quantification, traffic forecasting, and adaptive optimization. First, we develop a comprehensive model to assess the environmental, economic, and social consequences of climate-induced traffic disruptions under three top-priority climate change scenarios. Our analysis reveals significant welfare losses due to prolonged congestion and increased emissions under moderate to high climate change scenarios, underscoring the urgency of proactive adaptation. Building on this foundation, we present three key contributions. The first is AHSTN (Adaptive Hierarchical Spatio-Temporal Network), an end-to-end graph convolutional network that dynamically captures multiscale spatiotemporal dependencies in traffic data. By leveraging hierarchical learning mechanisms, AHSTN achieves superior prediction accuracy while reducing computational complexity, enabling real-time forecasting of weather-sensitive traffic patterns. The second contribution is prediction-driven signal control, a novel framework that integrates traffic flow forecasts with real-time traffic data to optimize signal timings. This method dynamically adjusts phase durations by incorporating both current and predicted vehicle queues, ensuring robustness under communication delays and uncertain weather conditions. The third contribution is CCDA (Centralized Critic and Decentralized Actors), a scalable RL architecture designed for low-frequency signal control. CCDA introduces the adjust all phases action design, allowing simultaneous optimization of signal durations across all phases in a cycle, thereby maximizing the impact of each intervention. Through decentralized actors and a centralized critic, CCDA balances localized decision-making with global coordination, ensuring adaptability to varying control frequencies while maintaining computational efficiency. Collectively, these advancements provide a scalable and adaptive approach to urban traffic management, bridging the gap between theoretical models and real-world operational requirements while enhancing resilience to climate change.
■590 ▼aSchool code: 0212.
■650 4▼aTraffic flow
■650 4▼aCommuting
■650 4▼aClimate change
■650 4▼aRain
■650 4▼aTraffic control
■650 4▼aMeteorology
■650 4▼aTransportation
■690 ▼a0404
■690 ▼a0543
■690 ▼a0557
■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=T17360850▼nKERIS▼z이 자료의 원문은 한국교육학술정보원에서 제공합니다.


