서브메뉴
검색
Data-Driven Predictive Control Beyond Linearity: An Autonomous Driving Perspective
Data-Driven Predictive Control Beyond Linearity: An Autonomous Driving Perspective
상세정보
- 자료유형
- 학위논문 서양
- 최종처리일시
- 20250211151456
- ISBN
- 9798384450092
- DDC
- 629.8
- 저자명
- Nair, Siddharth.
- 서명/저자
- Data-Driven Predictive Control Beyond Linearity: An Autonomous Driving Perspective
- 발행사항
- [Sl] : University of California, Berkeley, 2024
- 발행사항
- Ann Arbor : ProQuest Dissertations & Theses, 2024
- 형태사항
- 179 p
- 주기사항
- Source: Dissertations Abstracts International, Volume: 86-04, Section: B.
- 주기사항
- Advisor: Borrelli, Francesco.
- 학위논문주기
- Thesis (Ph.D.)--University of California, Berkeley, 2024.
- 초록/해제
- 요약Model Predictive Control (MPC) has been widely adopted in the industry for constrained optimal control, because of its straightforward transcription of the control problem and the availability of mature convex optimization solvers for efficient control synthesis. However as the name suggests, MPC requires a model of the underlying system and a quantification of the uncertainty in the assumed model for reliable performance and robustness. A vast body of work has been devoted to the analyses and design of robust and efficient MPC for linear systems with convex constraints. However, the world inherently involves nonlinear phenomena and non-convex decision-making, be it the dynamics of a bicycle, the multi-modality of a human driver, or the combinatorial problem of optimizing a route. While various MPC designs for nonlinear, non-convex settings have been studied, most suffer from at least one of the following shortcomings: (1) lack closed-loop performance guarantees or safety guarantees in the presence of uncertainty, (2) have small regions of feasibility due to conservative convexification of constraints or local linearization of dynamics, or (3) are intractable for real-time, high-frequency control.In this dissertation, we propose data-driven algorithms that exploit the nonlinearity and non-convexity in the optimal control problem for MPC designs that are robust to uncertainty and computationally efficient. The dissertation is divided into three parts focusing on robustness, performance, and computational efficiency. Each part presents algorithms that balance all three aspects, but are designed with an emphasis on the corresponding part's theme and demonstrated on practical applications in autonomous driving.
- 일반주제명
- Robotics
- 일반주제명
- Computer science
- 일반주제명
- Automotive engineering
- 일반주제명
- Mechanical engineering
- 키워드
- Machine learning
- 키워드
- Optimization
- 기타저자
- University of California, Berkeley Mechanical Engineering
- 기본자료저록
- Dissertations Abstracts International. 86-04B.
- 전자적 위치 및 접속
- 로그인 후 원문을 볼 수 있습니다.
MARC
008250123s2024 us c eng d■001000017161874
■00520250211151456
■006m o d
■007cr#unu||||||||
■020 ▼a9798384450092
■035 ▼a(MiAaPQ)AAI31297162
■040 ▼aMiAaPQ▼cMiAaPQ
■0820 ▼a629.8
■1001 ▼aNair, Siddharth.
■24510▼aData-Driven Predictive Control Beyond Linearity: An Autonomous Driving Perspective
■260 ▼a[Sl]▼bUniversity of California, Berkeley▼c2024
■260 1▼aAnn Arbor▼bProQuest Dissertations & Theses▼c2024
■300 ▼a179 p
■500 ▼aSource: Dissertations Abstracts International, Volume: 86-04, Section: B.
■500 ▼aAdvisor: Borrelli, Francesco.
■5021 ▼aThesis (Ph.D.)--University of California, Berkeley, 2024.
■520 ▼aModel Predictive Control (MPC) has been widely adopted in the industry for constrained optimal control, because of its straightforward transcription of the control problem and the availability of mature convex optimization solvers for efficient control synthesis. However as the name suggests, MPC requires a model of the underlying system and a quantification of the uncertainty in the assumed model for reliable performance and robustness. A vast body of work has been devoted to the analyses and design of robust and efficient MPC for linear systems with convex constraints. However, the world inherently involves nonlinear phenomena and non-convex decision-making, be it the dynamics of a bicycle, the multi-modality of a human driver, or the combinatorial problem of optimizing a route. While various MPC designs for nonlinear, non-convex settings have been studied, most suffer from at least one of the following shortcomings: (1) lack closed-loop performance guarantees or safety guarantees in the presence of uncertainty, (2) have small regions of feasibility due to conservative convexification of constraints or local linearization of dynamics, or (3) are intractable for real-time, high-frequency control.In this dissertation, we propose data-driven algorithms that exploit the nonlinearity and non-convexity in the optimal control problem for MPC designs that are robust to uncertainty and computationally efficient. The dissertation is divided into three parts focusing on robustness, performance, and computational efficiency. Each part presents algorithms that balance all three aspects, but are designed with an emphasis on the corresponding part's theme and demonstrated on practical applications in autonomous driving.
■590 ▼aSchool code: 0028.
■650 4▼aRobotics
■650 4▼aComputer science
■650 4▼aAutomotive engineering
■650 4▼aMechanical engineering
■653 ▼aAutonomous vehicles
■653 ▼aMachine learning
■653 ▼aNonlinear systems
■653 ▼aOptimization
■653 ▼aModel Predictive Control
■690 ▼a0771
■690 ▼a0984
■690 ▼a0548
■690 ▼a0540
■71020▼aUniversity of California, Berkeley▼bMechanical Engineering.
■7730 ▼tDissertations Abstracts International▼g86-04B.
■790 ▼a0028
■791 ▼aPh.D.
■792 ▼a2024
■793 ▼aEnglish
■85640▼uhttp://www.riss.kr/pdu/ddodLink.do?id=T17161874▼nKERIS▼z이 자료의 원문은 한국교육학술정보원에서 제공합니다.
